http://wikis.swarthmore.edu/ling073/api.php?action=feedcontributions&feedformat=atom&user=Jmundo1 2026-10-25T03:04:11Z User contributions MediaWiki 1.27.7 http://wikis.swarthmore.edu/ling073/index.php?title=User:Jmundo1/Final_project&diff=5381 2017-05-11T00:35:49Z

<p>Jmundo1: </p> <hr /> <div>=Transducer for Hiaki=<br /> <br /> For my final project, I created a prototype transducer for the Hiaki language, a language extremely lacking in computational resources. It is available at [https://github.swarthmore.edu/jmundo1/ling073-yaq Swarthmore&#039;s github] and publicly on [https://github.com/jdmundo/ling073-yaq-public github].<br /> <br /> ==Implementation==<br /> <br /> The implementation includes the following notable or unusual functionality, in addition to standard treatment of nouns, verbs, pronouns, postpositions, adjectives, adverbs, and other common Hiaki word types:<br /> <br /> ===Variation in Verb Forms===<br /> <br /> Hiaki verb forms exhibit somewhat idiosyncratic variation in their roots depending on whether derivational suffixation appears. This process could not be automated because of the idiosyncrasy, but the lexicon is built so both forms of a root verb can be placed and the transducer will be sure only to generate the right one.<br /> <br /> Examples:<br /> * poona (without morphology) to pon (with morphology)<br /> * kiima (without morphology) to kima&#039;a (with morphology)<br /> * ne&#039;e (without morphology) to ni&#039;i (with morphology)<br /> <br /> As the above examples illustrate, this change is unpredictable; it can involve addition of letters, change of letters, or deletion of letters. For this reason it must be hard-coded into the transducer.<br /> <br /> An important next step to the transducer would be to make sure both forms of all verbs are included; this would possibly require consultation with a speaker, as sources do not always include both forms.<br /> <br /> ===Noun Incorporation===<br /> <br /> Hiaki nouns and pronouns can be incorporated into verbs, losing their case morphology and becoming part of the single word. The transducer is structured to handle incorporation of arbitrary nouns into arbitrary verbs. <br /> <br /> Due to the computational complexity of noun incorporation, it is currently disabled. However, pronoun incorporation, which occurs vastly more frequently in the corpus, remains enabled.<br /> <br /> ===Reduplication===<br /> <br /> Many Hiaki verbs reduplicate productively to indicate, for instance, the habitual aspect on verbs, or plurality on adjectives. Functionality for reduplication is partially implemented using .twol rules. However, due to the complexity of Hiaki reduplication and the idiosyncracies across verbs regarding what part of the verb reduplicates, the implementation is not yet complete; in particular, it can only handle reduplication of single open or closed syllables, which must not begin with a vowel. (For example, the transducer can correctly analyze vivicha, the reduplicated form of vicha `see&#039;.)<br /> <br /> Some &quot;reduplications&quot; are not actually reduplications; these appear on verbs which have a different morphological change in place of reduplication for the same semantic effect. Such &quot;reduplications&quot;, which can be, for example, gemination of a consonant in the middle of the word, are hard-coded.<br /> <br /> Examples of Reduplication Types:<br /> *wiuta to wiwiuta (open syllable reduplication)<br /> *kupikte to kupikupikte (two open syllable reduplication)<br /> *hakta to hakhakta (closed syllable reduplication)<br /> *maveta to mavveta (gemination -- not actually reduplication)<br /> <br /> ===Spellrelax===<br /> <br /> A copy of the spellrelax file originally from [here http://svn.code.sf.net/p/apertium/svn/incubator/apertium-wal/dev/spellrelax.twol] is used to relax apostrophe-like symbols into a single symbol, as such symbols can appear inside Hiaki words.<br /> <br /> ==Next Steps==<br /> <br /> Several major grammatical aspects of the Hiaki language remain unimplemented, for example:<br /> <br /> ===Number Suppletion===<br /> <br /> Many Hiaki verbs completely change root form depending on the number of one of their arguments, usually the object if the verb is transitive and the subject otherwise. This is implemented in the transducer with &lt;sg&gt; and &lt;pl&gt; tags on verbs for those verbs known to exhibit number agreement; however, it is often the case in certain word databases that it is not made clear if a verb root is singular, plural, or both. More work needs to be done with the transducer to implement this properly.<br /> <br /> Also, currently the transducer has different lemmas for the two verbs, the singular and the plural form. To create any effective machine translation system in the future, it is likely these lemmas would need to be collapsed into one, which would require a small restructure of the transducer. For the purposes of the current transducer, the lemmas are kept distinct to avoid the confusion of a lemma with no resemblance to its form; for certain uses of the transducer, such as automating glossing, this is likely better. <br /> <br /> ===Vowel Change Rules===<br /> <br /> The vowel change rules in Hiaki are fairly complex, and the forms of affixes often vary in slight ways depending on the root. The transducer needs to be improved by first identifying the regularities and patterns in the alternation of affixes, then implementing them, probably through .twol. Currently some such patterns are implemented, but normally the transducer overgenerates, placing every form of a given affix on any root. There are also likely many forms not yet encountered which need to be put in the transducer.<br /> <br /> ===Root Change Rules===<br /> <br /> Many roots in Hiaki change form due to a variety of reasons, from number suppletion to the presence of a derivational suffix to being incorporated with another word to apparently idiosyncratic changes based on specific suffixes, such as wohoʼoria-po and wohoʼok-u, both of which mean `in a hole&#039; but which have slightly different roots in addition to different locative suffixes. Before the transducer can be improved to understand these changes, it must be better understood what exactly conditions them. This requires research into the work done on Hiaki and possibly consultation with speakers.<br /> <br /> ===Dealing with Spanish===<br /> <br /> A large amount of text available in Hiaki comes from transcribed interviews with native speakers, who often switch briefly into Spanish while speaking Hiaki. Several very common Spanish words are included in the transducer as loanwords, but there may be a better way to deal with switching into Spanish. This should be considered going forward in the transducer&#039;s development.<br /> <br /> ==Evaluation==<br /> <br /> Coverage was tested over a corpus (corpus/yaq.corpus.txt). Little text is readily available in Hiaki, so the corpus used only has 3254 words. To continue work on the transducer, it is likely a larger corpus would need to be compiled. <br /> <br /> Currently, the transducer tokenizes 4018 words in the corpus (including punctuation), and has coverage 67.21%.<br /> <br /> Precision and recall were evaluated on a small segment of 88 words of the corpus. The output of the transducer at the time of testing is stored, in CG format, in corpus/yaq.corpus.cg.txt; the correctly annotated version is stored in corpus/yaq.corpus.annotated.txt. Precision over this very small section was 80.53097%, and recall was 91.00000%. These high values were achieved because this is a well-understood section of the corpus (why it was chosen to be annotated), and thus the transducer is over-fit to it, having been built in reference to this part of the corpus. It was not feasible to measure precision and recall against a random part of the corpus, as the corpus is not well enough understood to be annotated.<br /> <br /> The transducer was then modified to improve precision and recall on yaq.corpus.cg.txt. The new output, in CG format, is storied in corpus/yaq.corpus.cg.new.txt. A new annotated version was made as well at corpus/yaq.corpus.cg.annotated.txt, as the transducer was also updated to parse quotation marks, which are now included in the new CG formatted documents. The new annotated corpus only differs from the old one in that it includes quotation marks. After the modification, precision was 74.07407% and recall was 95.23810%. The reason precision decreased may have to do with certain forms, which previously were not analyzed at all, now receiving multiple possibly correct analyses.<br /> <br /> [[Category:sp17_FinalProjects]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=User:Jmundo1/Final_project&diff=5380 2017-05-11T00:34:47Z

<p>Jmundo1: </p> <hr /> <div>=Transducer for Hiaki=<br /> <br /> For my final project, I created a prototype transducer for the Hiaki language, a language extremely lacking in computational resources. It is available at [https://github.swarthmore.edu/jmundo1/ling073-yaq Swarthmore&#039;s github] and publicly on [https://github.com/jdmundo/ling073-yaq-public github].<br /> <br /> ==Implementation==<br /> <br /> The implementation includes the following notable or unusual functionality, in addition to standard treatment of nouns, verbs, pronouns, postpositions, adjectives, adverbs, and other common Hiaki word types:<br /> <br /> ===Variation in Verb Forms===<br /> <br /> Hiaki verb forms exhibit somewhat idiosyncratic variation in their roots depending on whether derivational suffixation appears. This process could not be automated because of the idiosyncrasy, but the lexicon is built so both forms of a root verb can be placed and the transducer will be sure only to generate the right one.<br /> <br /> Examples:<br /> * poona (without morphology) to pon (with morphology)<br /> * kiima (without morphology) to kima&#039;a (with morphology)<br /> * ne&#039;e (without morphology) to ni&#039;i (with morphology)<br /> <br /> As the above examples illustrate, this change is unpredictable; it can involve addition of letters, change of letters, or deletion of letters. For this reason it must be hard-coded into the transducer.<br /> <br /> An important next step to the transducer would be to make sure both forms of all verbs are included; this would possibly require consultation with a speaker, as sources do not always include both forms.<br /> <br /> ===Noun Incorporation===<br /> <br /> Hiaki nouns and pronouns can be incorporated into verbs, losing their case morphology and becoming part of the single word. The transducer is structured to handle incorporation of arbitrary nouns into arbitrary verbs. <br /> <br /> Due to the computational complexity of noun incorporation, it is currently disabled. However, pronoun incorporation, which occurs vastly more frequently in the corpus, remains enabled.<br /> <br /> ===Reduplication===<br /> <br /> Many Hiaki verbs reduplicate productively to indicate, for instance, the habitual aspect on verbs, or plurality on adjectives. Functionality for reduplication is partially implemented using .twol rules. However, due to the complexity of Hiaki reduplication and the idiosyncracies across verbs regarding what part of the verb reduplicates, the implementation is not yet complete; in particular, it can only handle reduplication of single open or closed syllables, which must not begin with a vowel. (For example, the transducer can correctly analyze vivicha, the reduplicated form of vicha `see&#039;.)<br /> <br /> Some &quot;reduplications&quot; are not actually reduplications; these appear on verbs which have a different morphological change in place of reduplication for the same semantic effect. Such &quot;reduplications&quot;, which can be, for example, gemination of a consonant in the middle of the word, are hard-coded.<br /> <br /> Examples of Reduplication Types:<br /> *wiuta to wiwiuta (open syllable reduplication)<br /> *kupikte to kupikupikte (two open syllable reduplication)<br /> *hakta to hakhakta (closed syllable reduplication)<br /> *maveta to mavveta (gemination -- not actually reduplication)<br /> <br /> ===Spellrelax===<br /> <br /> A copy of the spellrelax file originally from [here http://svn.code.sf.net/p/apertium/svn/incubator/apertium-wal/dev/spellrelax.twol] is used to relax apostrophe-like symbols into a single symbol, as such symbols can appear inside Hiaki words.<br /> <br /> ==Next Steps==<br /> <br /> Several major grammatical aspects of the Hiaki language remain unimplemented, for example:<br /> <br /> ===Number Suppletion===<br /> <br /> Many Hiaki verbs completely change root form depending on the number of one of their arguments, usually the object if the verb is transitive and the subject otherwise. This is implemented in the transducer with &lt;sg&gt; and &lt;pl&gt; tags on verbs for those verbs known to exhibit number agreement; however, it is often the case in certain word databases that it is not made clear if a verb root is singular, plural, or both. More work needs to be done with the transducer to implement this properly.<br /> <br /> Also, currently the transducer has different lemmas for the two verbs, the singular and the plural form. To create any effective machine translation system in the future, it is likely these lemmas would need to be collapsed into one, which would require a small restructure of the transducer. For the purposes of the current transducer, the lemmas are kept distinct to avoid the confusion of a lemma with no resemblance to its form; for certain uses of the transducer, such as automating glossing, this is likely better. <br /> <br /> ===Vowel Change Rules===<br /> <br /> The vowel change rules in Hiaki are fairly complex, and the forms of affixes often vary in slight ways depending on the root. The transducer needs to be improved by first identifying the regularities and patterns in the alternation of affixes, then implementing them, probably through .twol. Currently some such patterns are implemented, but normally the transducer overgenerates, placing every form of a given affix on any root. There are also likely many forms not yet encountered which need to be put in the transducer.<br /> <br /> ===Root Change Rules===<br /> <br /> Many roots in Hiaki change form due to a variety of reasons, from number suppletion to the presence of a derivational suffix to being incorporated with another word to apparently idiosyncratic changes based on specific suffixes, such as wohoʼoria-po and wohoʼok-u, both of which mean `in a hole&#039; but which have slightly different roots in addition to different locative suffixes. Before the transducer can be improved to understand these changes, it must be better understood what exactly conditions them. This requires research into the work done on Hiaki and possibly consultation with speakers.<br /> <br /> ===Dealing with Spanish===<br /> <br /> A large amount of text available in Hiaki comes from transcribed interviews with native speakers, who often switch briefly into Spanish while speaking Hiaki. Several very common Spanish words are included in the transducer as loanwords, but there may be a better way to deal with switching into Spanish. This should be considered going forward in the transducer&#039;s development.<br /> <br /> ==Evaluation==<br /> <br /> Coverage was tested over a corpus (corpus/yaq.corpus.txt). Little text is readily available in Hiaki, so the corpus used only has 3254 words. To continue work on the transducer, it is likely a larger corpus would need to be compiled. <br /> <br /> Currently, the transducer tokenizes 4018 words in the corpus (including punctuation), and has coverage 67.21%.<br /> <br /> Precision and recall were evaluated on a small segment of 88 words of the corpus. The output of the transducer at the time of testing is stored, in CG format, in corpus/yaq.corpus.cg.txt; the correctly annotated version is stored in corpus/yaq.corpus.annotated.txt. Precision over this very small section was 80.53097%, and recall was 91.00000%. These high values were achieved because this is a well-understood section of the corpus (why it was chosen to be annotated), and thus the transducer is over-fit to it, having been built in reference to this part of the corpus.<br /> <br /> The transducer was then modified to improve precision and recall on yaq.corpus.cg.txt. The new output, in CG format, is storied in corpus/yaq.corpus.cg.new.txt. A new annotated version was made as well at corpus/yaq.corpus.cg.annotated.txt, as the transducer was also updated to parse quotation marks, which are now included in the new CG formatted documents. The new annotated corpus only differs from the old one in that it includes quotation marks. After the modification, precision was 74.07407% and recall was 95.23810%. The reason precision decreased may have to do with certain forms, which previously were not analyzed at all, now receiving multiple possibly correct analyses.<br /> <br /> [[Category:sp17_FinalProjects]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=User:Jmundo1/Final_project&diff=5376 2017-05-11T00:11:24Z

<p>Jmundo1: </p> <hr /> <div>=Transducer for Hiaki=<br /> <br /> For my final project, I created a prototype transducer for the Hiaki language, a language extremely lacking in computational resources. It is available at [https://github.swarthmore.edu/jmundo1/ling073-yaq Swarthmore&#039;s github].<br /> <br /> ==Implementation==<br /> <br /> The implementation includes the following notable or unusual functionality, in addition to standard treatment of nouns, verbs, pronouns, postpositions, adjectives, adverbs, and other common Hiaki word types:<br /> <br /> ===Variation in Verb Forms===<br /> <br /> Hiaki verb forms exhibit somewhat idiosyncratic variation in their roots depending on whether derivational suffixation appears. This process could not be automated because of the idiosyncrasy, but the lexicon is built so both forms of a root verb can be placed and the transducer will be sure only to generate the right one.<br /> <br /> Examples:<br /> * poona (without morphology) to pon (with morphology)<br /> * kiima (without morphology) to kima&#039;a (with morphology)<br /> * ne&#039;e (without morphology) to ni&#039;i (with morphology)<br /> <br /> As the above examples illustrate, this change is unpredictable; it can involve addition of letters, change of letters, or deletion of letters. For this reason it must be hard-coded into the transducer.<br /> <br /> An important next step to the transducer would be to make sure both forms of all verbs are included; this would possibly require consultation with a speaker, as sources do not always include both forms.<br /> <br /> ===Noun Incorporation===<br /> <br /> Hiaki nouns and pronouns can be incorporated into verbs, losing their case morphology and becoming part of the single word. The transducer is structured to handle incorporation of arbitrary nouns into arbitrary verbs. <br /> <br /> Due to the computational complexity of noun incorporation, it is currently disabled. However, pronoun incorporation, which occurs vastly more frequently in the corpus, remains enabled.<br /> <br /> ===Reduplication===<br /> <br /> Many Hiaki verbs reduplicate productively to indicate, for instance, the habitual aspect on verbs, or plurality on adjectives. Functionality for reduplication is partially implemented using .twol rules. However, due to the complexity of Hiaki reduplication and the idiosyncracies across verbs regarding what part of the verb reduplicates, the implementation is not yet complete; in particular, it can only handle reduplication of single open or closed syllables, which must not begin with a vowel. (For example, the transducer can correctly analyze vivicha, the reduplicated form of vicha `see&#039;.)<br /> <br /> Some &quot;reduplications&quot; are not actually reduplications; these appear on verbs which have a different morphological change in place of reduplication for the same semantic effect. Such &quot;reduplications&quot;, which can be, for example, gemination of a consonant in the middle of the word, are hard-coded.<br /> <br /> Examples of Reduplication Types:<br /> *wiuta to wiwiuta (open syllable reduplication)<br /> *kupikte to kupikupikte (two open syllable reduplication)<br /> *hakta to hakhakta (closed syllable reduplication)<br /> *maveta to mavveta (gemination -- not actually reduplication)<br /> <br /> ===Spellrelax===<br /> <br /> A copy of the spellrelax file originally from [here http://svn.code.sf.net/p/apertium/svn/incubator/apertium-wal/dev/spellrelax.twol] is used to relax apostrophe-like symbols into a single symbol, as such symbols can appear inside Hiaki words.<br /> <br /> ==Next Steps==<br /> <br /> Several major grammatical aspects of the Hiaki language remain unimplemented, for example:<br /> <br /> ===Number Suppletion===<br /> <br /> Many Hiaki verbs completely change root form depending on the number of one of their arguments, usually the object if the verb is transitive and the subject otherwise. This is implemented in the transducer with &lt;sg&gt; and &lt;pl&gt; tags on verbs for those verbs known to exhibit number agreement; however, it is often the case in certain word databases that it is not made clear if a verb root is singular, plural, or both. More work needs to be done with the transducer to implement this properly.<br /> <br /> Also, currently the transducer has different lemmas for the two verbs, the singular and the plural form. To create any effective machine translation system in the future, it is likely these lemmas would need to be collapsed into one, which would require a small restructure of the transducer. For the purposes of the current transducer, the lemmas are kept distinct to avoid the confusion of a lemma with no resemblance to its form; for certain uses of the transducer, such as automating glossing, this is likely better. <br /> <br /> ===Vowel Change Rules===<br /> <br /> The vowel change rules in Hiaki are fairly complex, and the forms of affixes often vary in slight ways depending on the root. The transducer needs to be improved by first identifying the regularities and patterns in the alternation of affixes, then implementing them, probably through .twol. Currently some such patterns are implemented, but normally the transducer overgenerates, placing every form of a given affix on any root. There are also likely many forms not yet encountered which need to be put in the transducer.<br /> <br /> ===Root Change Rules===<br /> <br /> Many roots in Hiaki change form due to a variety of reasons, from number suppletion to the presence of a derivational suffix to being incorporated with another word to apparently idiosyncratic changes based on specific suffixes, such as wohoʼoria-po and wohoʼok-u, both of which mean `in a hole&#039; but which have slightly different roots in addition to different locative suffixes. Before the transducer can be improved to understand these changes, it must be better understood what exactly conditions them. This requires research into the work done on Hiaki and possibly consultation with speakers.<br /> <br /> ===Dealing with Spanish===<br /> <br /> A large amount of text available in Hiaki comes from transcribed interviews with native speakers, who often switch briefly into Spanish while speaking Hiaki. Several very common Spanish words are included in the transducer as loanwords, but there may be a better way to deal with switching into Spanish. This should be considered going forward in the transducer&#039;s development.<br /> <br /> ==Evaluation==<br /> <br /> Coverage was tested over a corpus (corpus/yaq.corpus.txt). Little text is readily available in Hiaki, so the corpus used only has 3254 words. To continue work on the transducer, it is likely a larger corpus would need to be compiled. <br /> <br /> Currently, the transducer tokenizes 4018 words in the corpus (including punctuation), and has coverage 67.21%.<br /> <br /> Precision and recall were evaluated on a small segment of 88 words of the corpus. The output of the transducer at the time of testing is stored, in CG format, in corpus/yaq.corpus.cg.txt; the correctly annotated version is stored in corpus/yaq.corpus.annotated.txt. Precision over this very small section was 80.53097%, and recall was 91.00000%. These high values were achieved because this is a well-understood section of the corpus (why it was chosen to be annotated), and thus the transducer is over-fit to it, having been built in reference to this part of the corpus.<br /> <br /> The transducer was then modified to improve precision and recall on yaq.corpus.cg.txt. The new output, in CG format, is storied in corpus/yaq.corpus.cg.new.txt. A new annotated version was made as well at corpus/yaq.corpus.cg.annotated.txt, as the transducer was also updated to parse quotation marks, which are now included in the new CG formatted documents. The new annotated corpus only differs from the old one in that it includes quotation marks. After the modification, precision was 74.07407% and recall was 95.23810%. The reason precision decreased may have to do with certain forms, which previously were not analyzed at all, now receiving multiple possibly correct analyses.<br /> <br /> [[Category:sp17_FinalProjects]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=User:Jmundo1/Final_project&diff=5374 2017-05-11T00:07:30Z

<p>Jmundo1: </p> <hr /> <div>=Transducer for Hiaki=<br /> <br /> For my final project, I created a prototype transducer for the Hiaki language, a language extremely lacking in computational resources. It is available at [https://github.swarthmore.edu/jmundo1/ling073-yaq Swarthmore&#039;s github].<br /> <br /> ==Implementation==<br /> <br /> The implementation includes the following notable or unusual functionality, in addition to standard treatment of nouns, verbs, pronouns, postpositions, adjectives, adverbs, and other common Hiaki word types:<br /> <br /> ===Variation in Verb Forms===<br /> <br /> Hiaki verb forms exhibit somewhat idiosyncratic variation in their roots depending on whether derivational suffixation appears. This process could not be automated because of the idiosyncrasy, but the lexicon is built so both forms of a root verb can be placed and the transducer will be sure only to generate the right one.<br /> <br /> Examples:<br /> * poona (without morphology) to pon (with morphology)<br /> * kiima (without morphology) to kima&#039;a (with morphology)<br /> * ne&#039;e (without morphology) to ni&#039;i (with morphology)<br /> <br /> As the above examples illustrate, this change is unpredictable; it can involve addition of letters, change of letters, or deletion of letters. For this reason it must be hard-coded into the transducer.<br /> <br /> An important next step to the transducer would be to make sure both forms of all verbs are included; this would possibly require consultation with a speaker, as sources do not always include both forms.<br /> <br /> ===Noun Incorporation===<br /> <br /> Hiaki nouns and pronouns can be incorporated into verbs, losing their case morphology and becoming part of the single word. The transducer is structured to handle incorporation of arbitrary nouns into arbitrary verbs. <br /> <br /> Due to the computational complexity of noun incorporation, it is currently disabled. However, pronoun incorporation, which occurs vastly more frequently in the corpus, remains enabled.<br /> <br /> ===Reduplication===<br /> <br /> Many Hiaki verbs reduplicate productively to indicate, for instance, the habitual aspect on verbs, or plurality on adjectives. Functionality for reduplication is partially implemented using .twol rules. However, due to the complexity of Hiaki reduplication and the idiosyncracies across verbs regarding what part of the verb reduplicates, the implementation is not yet complete; in particular, it can only handle reduplication of single open or closed syllables, which must not begin with a vowel. (For example, the transducer can correctly analyze vivicha, the reduplicated form of vicha `see&#039;.)<br /> <br /> Some &quot;reduplications&quot; are not actually reduplications; these appear on verbs which have a different morphological change in place of reduplication for the same semantic effect. Such &quot;reduplications&quot;, which can be, for example, gemination of a consonant in the middle of the word, are hard-coded.<br /> <br /> Examples of Reduplication Types:<br /> *wiuta to wiwiuta (open syllable reduplication)<br /> *kupikte to kupikupikte (two open syllable reduplication)<br /> *hakta to hakhakta (closed syllable reduplication)<br /> *maveta to mavveta (gemination -- not actually reduplication)<br /> <br /> ===Spellrelax===<br /> <br /> A copy of the spellrelax file originally from [here http://svn.code.sf.net/p/apertium/svn/incubator/apertium-wal/dev/spellrelax.twol] is used to relax apostrophe-like symbols into a single symbol, as such symbols can appear inside Hiaki words.<br /> <br /> ==Next Steps==<br /> <br /> Several major grammatical aspects of the Hiaki language remain unimplemented, for example:<br /> <br /> ===Number Suppletion===<br /> <br /> Many Hiaki verbs completely change root form depending on the number of one of their arguments, usually the object if the verb is transitive and the subject otherwise. This is implemented in the transducer with &lt;sg&gt; and &lt;pl&gt; tags on verbs for those verbs known to exhibit number agreement; however, it is often the case in certain word databases that it is not made clear if a verb root is singular, plural, or both. More work needs to be done with the transducer to implement this properly.<br /> <br /> Also, currently the transducer has different lemmas for the two verbs, the singular and the plural form. To create any effective machine translation system in the future, it is likely these lemmas would need to be collapsed into one, which would require a small restructure of the transducer. For the purposes of the current transducer, the lemmas are kept distinct to avoid the confusion of a lemma with no resemblance to its form; for certain uses of the transducer, such as automating glossing, this is likely better. <br /> <br /> ===Vowel Change Rules===<br /> <br /> The vowel change rules in Hiaki are fairly complex, and the forms of affixes often vary in slight ways depending on the root. The transducer needs to be improved by first identifying the regularities and patterns in the alternation of affixes, then implementing them, probably through .twol. Currently some such patterns are implemented, but normally the transducer overgenerates, placing every form of a given affix on any root. There are also likely many forms not yet encountered which need to be put in the transducer.<br /> <br /> ===Root Change Rules===<br /> <br /> Many roots in Hiaki change form due to a variety of reasons, from number suppletion to the presence of a derivational suffix to being incorporated with another word to apparently idiosyncratic changes based on specific suffixes, such as wohoʼoria-po and wohoʼok-u, both of which mean `in a hole&#039; but which have slightly different roots in addition to different locative suffixes. Before the transducer can be improved to understand these changes, it must be better understood what exactly conditions them. This requires research into the work done on Hiaki and possibly consultation with speakers.<br /> <br /> ===Dealing with Spanish===<br /> <br /> A large amount of text available in Hiaki comes from transcribed interviews with native speakers, who often switch briefly into Spanish while speaking Hiaki. Several very common Spanish words are included in the transducer as loanwords, but there may be a better way to deal with switching into Spanish. This should be considered going forward in the transducer&#039;s development.<br /> <br /> ==Evaluation==<br /> <br /> Coverage was tested over a corpus (corpus/yaq.corpus.txt). Little text is readily available in Hiaki, so the corpus used only has 3254 words. To continue work on the transducer, it is likely a larger corpus would need to be compiled. <br /> <br /> Currently, the transducer tokenizes 4018 words in the corpus (including punctuation), and has coverage 67.21%.<br /> <br /> Precision and recall were evaluated on a small segment of 88 words of the corpus. The output of the transducer at the time of testing is stored, in CG format, in corpus/yaq.corpus.cg.txt; the correctly annotated version is stored in corpus/yaq.corpus.annotated.txt. Precision over this very small section is 80.53097%, and recall is 91.00000%. These high values are achieved because this is a well-understood section of the corpus (why it was chosen to be annotated), and thus the transducer is over-fit to it, having been built in reference to this part of the corpus.<br /> <br /> The transducer was then modified to improve precision and recall on yaq.corpus.cg.txt. The new output, in CG format, is storied in corpus/yaq.corpus.cg.new.txt. A new annotated version was made as well at corpus/yaq.corpus.cg.annotated.txt, as the transducer was also updated to parse quotation marks, which are now included in the new CG formatted documents. The new annotated corpus only differs from the old one in that it includes quotation marks. After the modification, precision was 74.07407% and recall was 95.23810%. The reason precision decreased may have to do with certain forms, which previously were not analyzed at all, now receiving multiple possibly correct analyses.<br /> <br /> [[Category:sp17_FinalProjects]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=User:Jmundo1/Final_project&diff=5373 2017-05-11T00:06:14Z

<p>Jmundo1: </p> <hr /> <div>=Transducer for Hiaki=<br /> <br /> For my final project, I created a prototype transducer for the Hiaki language, a language extremely lacking in computational resources. It is available at [https://github.swarthmore.edu/jmundo1/ling073-yaq Swarthmore&#039;s github].<br /> <br /> ==Implementation==<br /> <br /> The implementation includes the following notable or unusual functionality, in addition to standard treatment of nouns, verbs, pronouns, postpositions, adjectives, adverbs, and other common Hiaki word types:<br /> <br /> ===Variation in Verb Forms===<br /> <br /> Hiaki verb forms exhibit somewhat idiosyncratic variation in their roots depending on whether derivational suffixation appears. This process could not be automated because of the idiosyncrasy, but the lexicon is built so both forms of a root verb can be placed and the transducer will be sure only to generate the right one.<br /> <br /> Examples:<br /> * poona (without morphology) to pon (with morphology)<br /> * kiima (without morphology) to kima&#039;a (with morphology)<br /> * ne&#039;e (without morphology) to ni&#039;i (with morphology)<br /> <br /> As the above examples illustrate, this change is unpredictable; it can involve addition of letters, change of letters, or deletion of letters. For this reason it must be hard-coded into the transducer.<br /> <br /> An important next step to the transducer would be to make sure both forms of all verbs are included; this would possibly require consultation with a speaker, as sources do not always include both forms.<br /> <br /> ===Noun Incorporation===<br /> <br /> Hiaki nouns and pronouns can be incorporated into verbs, losing their case morphology and becoming part of the single word. The transducer is structured to handle incorporation of arbitrary nouns into arbitrary verbs. <br /> <br /> Due to the computational complexity of noun incorporation, it is currently disabled. However, pronoun incorporation, which occurs vastly more frequently in the corpus, remains enabled.<br /> <br /> ===Reduplication===<br /> <br /> Many Hiaki verbs reduplicate productively to indicate, for instance, the habitual aspect on verbs, or plurality on adjectives. Functionality for reduplication is partially implemented using .twol rules. However, due to the complexity of Hiaki reduplication and the idiosyncracies across verbs regarding what part of the verb reduplicates, the implementation is not yet complete; in particular, it can only handle reduplication of single open or closed syllables, which must not begin with a vowel. (For example, the transducer can correctly analyze vivicha, the reduplicated form of vicha `see&#039;.)<br /> <br /> Some &quot;reduplications&quot; are not actually reduplications; these appear on verbs which have a different morphological change in place of reduplication for the same semantic effect. Such &quot;reduplications&quot;, which can be, for example, gemination of a consonant in the middle of the word, are hard-coded.<br /> <br /> Examples of Reduplication Types:<br /> *wiuta to wiwiuta (open syllable reduplication)<br /> *kupikte to kupikupikte (two open syllable reduplication)<br /> *hakta to hakhakta (closed syllable reduplication)<br /> *maveta to mavveta (gemination -- not actually reduplication)<br /> <br /> ===Spellrelax===<br /> <br /> A copy of the spellrelax file originally from [here http://svn.code.sf.net/p/apertium/svn/incubator/apertium-wal/dev/spellrelax.twol] is used to relax apostrophe-like symbols into a single symbol, as such symbols can appear inside Hiaki words.<br /> <br /> ==Next Steps==<br /> <br /> Several major grammatical aspects of the Hiaki language remain unimplemented, for example:<br /> <br /> ===Number Suppletion===<br /> <br /> Many Hiaki verbs completely change root form depending on the number of one of their arguments, usually the object if the verb is transitive and the subject otherwise. This is implemented in the transducer with &lt;sg&gt; and &lt;pl&gt; tags on verbs for those verbs known to exhibit number agreement; however, it is often the case in certain word databases that it is not made clear if a verb root is singular, plural, or both. More work needs to be done with the transducer to implement this properly.<br /> <br /> Also, currently the transducer has different lemmas for the two verbs, the singular and the plural form. To create any effective machine translation system in the future, it is likely these lemmas would need to be collapsed into one, which would require a small restructure of the transducer. For the purposes of the current transducer, the lemmas are kept distinct to avoid the confusion of a lemma with no resemblance to its form; for certain uses of the transducer, such as automating glossing, this is likely better. <br /> <br /> ===Vowel Change Rules===<br /> <br /> The vowel change rules in Hiaki are fairly complex, and the forms of affixes often vary in slight ways depending on the root. The transducer needs to be improved by first identifying the regularities and patterns in the alternation of affixes, then implementing them, probably through .twol. Currently some such patterns are implemented, but normally the transducer overgenerates, placing every form of a given affix on any root. There are also likely forms we have not encountered to put in the transducer.<br /> <br /> ===Root Change Rules===<br /> <br /> Many roots in Hiaki change form due to a variety of reasons, from number suppletion to the presence of a derivational suffix to being incorporated with another word to apparently idiosyncratic changes based on specific suffixes, such as wohoʼoria-po and wohoʼok-u, both of which mean `in a hole&#039; but which have slightly different roots in addition to different locative suffixes. Before the transducer can be improved to understand these changes, it must be better understood what exactly conditions them. This requires research into the work done on Hiaki and possibly consultation with speakers.<br /> <br /> ===Dealing with Spanish===<br /> <br /> A large amount of text available in Hiaki comes from transcribed interviews with native speakers, who often switch briefly into Spanish while speaking Hiaki. Several very common Spanish words are included in the transducer as loanwords, but there may be a better way to deal with switching into Spanish. This should be considered going forward in the transducer&#039;s development.<br /> <br /> ==Evaluation==<br /> <br /> Coverage was tested over a corpus (corpus/yaq.corpus.txt). Little text is readily available in Hiaki, so the corpus used only has 3254 words. To continue work on the transducer, it is likely a larger corpus would need to be compiled. <br /> <br /> Currently, the transducer tokenizes 4018 words in the corpus (including punctuation), and has coverage 67.21%.<br /> <br /> Precision and recall were evaluated on a small segment of 88 words of the corpus. The output of the transducer at the time of testing is stored, in CG format, in corpus/yaq.corpus.cg.txt; the correctly annotated version is stored in corpus/yaq.corpus.annotated.txt. Precision over this very small section is 80.53097%, and recall is 91.00000%. These high values are achieved because this is a well-understood section of the corpus (why it was chosen to be annotated), and thus the transducer is over-fit to it, having been built in reference to this part of the corpus.<br /> <br /> The transducer was then modified to improve precision and recall on yaq.corpus.cg.txt. The new output, in CG format, is storied in corpus/yaq.corpus.cg.new.txt. A new annotated version was made as well at corpus/yaq.corpus.cg.annotated.txt, as the transducer was also updated to parse quotation marks, which are now included in the new CG formatted documents. The new annotated corpus only differs from the old one in that it includes quotation marks. After the modification, precision was 74.07407% and recall was 95.23810%. The reason precision decreased may have to do with certain forms, which previously were not analyzed at all, now receiving multiple possibly correct analyses.<br /> <br /> [[Category:sp17_FinalProjects]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=User:Jmundo1/Final_project&diff=5372 2017-05-11T00:05:44Z

<p>Jmundo1: </p> <hr /> <div>=Transducer for Hiaki=<br /> <br /> For my final project, I created a prototype transducer for the Hiaki language, a language extremely lacking in computational resources. It is available at [https://github.swarthmore.edu/jmundo1/ling073-yaq Swarthmore&#039;s github].<br /> <br /> ==Implementation==<br /> <br /> The implementation includes the following notable or unusual functionality, in addition to standard treatment of nouns, verbs, pronouns, postpositions, adjectives, adverbs, and other common Hiaki word types:<br /> <br /> ===Variation in Verb Forms===<br /> <br /> Hiaki verb forms exhibit somewhat idiosyncratic variation in their roots depending on whether derivational suffixation appears. This process could not be automated because of the idiosyncrasy, but the lexicon is built so both forms of a root verb can be placed and the transducer will be sure only to generate the right one.<br /> <br /> Examples:<br /> * poona (without morphology) to pon (with morphology)<br /> * kiima (without morphology) to kima&#039;a (with morphology)<br /> * ne&#039;e (without morphology) to ni&#039;i (with morphology)<br /> <br /> As the above examples illustrate, this change is unpredictable; it can involve addition of letters, change of letters, or deletion of letters. For this reason it must be hard-coded into the transducer.<br /> <br /> An important next step to the transducer would be to make sure both forms of all verbs are included; this would possibly require consultation with a speaker, as sources do not always include both forms.<br /> <br /> ===Noun Incorporation===<br /> <br /> Hiaki nouns and pronouns can be incorporated into verbs, losing their case morphology and becoming part of the single word. The transducer is structured to handle incorporation of arbitrary nouns into arbitrary verbs. <br /> <br /> Due to the computational complexity of noun incorporation, it is currently disabled. However, pronoun incorporation, which occurs vastly more frequently in the corpus, remains enabled.<br /> <br /> ===Reduplication===<br /> <br /> Many Hiaki verbs reduplicate productively to indicate, for instance, the habitual aspect on verbs, or plurality on adjectives. Functionality for reduplication is partially implemented using .twol rules. However, due to the complexity of Hiaki reduplication and the idiosyncracies across verbs regarding what part of the verb reduplicates, the implementation is not yet complete; in particular, it can only handle reduplication of single open or closed syllables, which must not begin with a vowel. (For example, the transducer can correctly analyze vivicha, the reduplicated form of vicha `see&#039;.)<br /> <br /> Some &quot;reduplications&quot; are not actually reduplications; these appear on verbs which have a different morphological change in place of reduplication for the same semantic effect. Such &quot;reduplications&quot;, which can be, for example, gemination of a consonant in the middle of the word, are hard-coded.<br /> <br /> Examples of Reduplication Types:<br /> *wiuta to wiwiuta (open syllable reduplication)<br /> *kupikte to kupikupikte (two open syllable reduplication)<br /> *hakta to hakhakta (closed syllable reduplication)<br /> *maveta to mavveta (gemination -- not actually reduplication)<br /> <br /> ===Spellrelax===<br /> <br /> A copy of the spellrelax file originally from [here http://svn.code.sf.net/p/apertium/svn/incubator/apertium-wal/dev/spellrelax.twol] is used to relax apostrophe-like symbols into a single symbol, as such symbols can appear inside Hiaki words.<br /> <br /> ==Next Steps==<br /> <br /> Several major grammatical aspects of the Hiaki language remain unimplemented, for example:<br /> <br /> ===Number Suppletion===<br /> <br /> Many Hiaki verbs completely change root form depending on the number of one of their arguments, usually the object if the verb is transitive and the subject otherwise. This is implemented in the transducer with &lt;sg&gt; and &lt;pl&gt; tags on verbs for those verbs known to exhibit number agreement; however, it is often the case in certain word databases that it is not made clear if a verb root is singular, plural, or both. More work needs to be done with the transducer to implement this properly.<br /> <br /> Also, currently the transducer has different lemmas for the two verbs, the singular and the plural form. To create any effective machine translation system in the future, it is likely these lemmas would need to be collapsed into one, which would require a small restructure of the transducer. For the purposes of the current transducer, the lemmas are kept distinct to avoid the confusion of a lemma with no resemblance to its form; for certain uses of the transducer, such as automating glossing, this is likely better. <br /> <br /> ===Vowel Change Rules===<br /> <br /> The vowel change rules in Hiaki are fairly complex, and the forms of affixes often vary in slight ways depending on the root. The transducer needs to be improved by first identifying the regularities and patterns in the alternation of affixes, then implementing them, probably through .twol. Currently some such patterns are implemented, but normally the transducer overgenerates, placing every form of a given affix on any root. There are also likely forms we have not encountered to put in the transducer.<br /> <br /> ===Root Change Rules===<br /> <br /> Many roots in Hiaki change form due to a variety of reasons, from number suppletion to the presence of a derivational suffix to being incorporated with another word to apparently idiosyncratic changes based on specific suffixes, such as wohoʼoria-po and wohoʼok-u, both of which mean `in a hole&#039; but which have slightly different roots in addition to different locative suffixes. Before the transducer can be improved to understand these changes, it must be better understood what exactly conditions them. This requires research into the work done on Hiaki and possibly consultation with speakers.<br /> <br /> ===Dealing with Spanish===<br /> <br /> A large amount of text available in Hiaki comes from transcribed interviews with native speakers, who often switch briefly into Spanish while speaking Hiaki. Several very common Spanish words are included in the transducer as loanwords, but there may be a better way to deal with switching into Spanish. This should be considered going forward in the transducer&#039;s development.<br /> <br /> ==Evaluation==<br /> <br /> Coverage was tested over a corpus (corpus/yaq.corpus.txt). Little text is readily available in Hiaki, so the corpus used only has 3254 words. To continue work on the transducer, it is likely a larger corpus would need to be compiled. <br /> <br /> Currently, the transducer tokenizes 4018 words in the corpus (including punctuation), and has coverage 66.48%.<br /> <br /> Precision and recall were evaluated on a small segment of 88 words of the corpus. The output of the transducer at the time of testing is stored, in CG format, in corpus/yaq.corpus.cg.txt; the correctly annotated version is stored in corpus/yaq.corpus.annotated.txt. Precision over this very small section is 80.53097%, and recall is 91.00000%. These high values are achieved because this is a well-understood section of the corpus (why it was chosen to be annotated), and thus the transducer is over-fit to it, having been built in reference to this part of the corpus.<br /> <br /> The transducer was then modified to improve precision and recall on yaq.corpus.cg.txt. The new output, in CG format, is storied in corpus/yaq.corpus.cg.new.txt. A new annotated version was made as well at corpus/yaq.corpus.cg.annotated.txt, as the transducer was also updated to parse quotation marks, which are now included in the new CG formatted documents. The new annotated corpus only differs from the old one in that it includes quotation marks. After the modification, precision was 74.07407% and recall was 95.23810%. The reason precision decreased may have to do with certain forms, which previously were not analyzed at all, now receiving multiple possibly correct analyses.<br /> <br /> [[Category:sp17_FinalProjects]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=User:Jmundo1/Final_project&diff=5371 2017-05-11T00:03:35Z

<p>Jmundo1: </p> <hr /> <div>=Transducer for Hiaki=<br /> <br /> For my final project, I created a prototype transducer for the Hiaki language, a language extremely lacking in computational resources. It is available at [https://github.swarthmore.edu/jmundo1/ling073-yaq Swarthmore&#039;s github].<br /> <br /> ==Implementation==<br /> <br /> The implementation includes the following notable or unusual functionality, in addition to standard treatment of nouns, verbs, pronouns, postpositions, adjectives, adverbs, and other common Hiaki word types:<br /> <br /> ===Variation in Verb Forms===<br /> <br /> Hiaki verb forms exhibit somewhat idiosyncratic variation in their roots depending on whether derivational suffixation appears. This process could not be automated because of the idiosyncrasy, but the lexicon is built so both forms of a root verb can be placed and the transducer will be sure only to generate the right one.<br /> <br /> Examples:<br /> * poona (without morphology) to pon (with morphology)<br /> * kiima (without morphology) to kima&#039;a (with morphology)<br /> * ne&#039;e (without morphology) to ni&#039;i (with morphology)<br /> <br /> As the above examples illustrate, this change is unpredictable; it can involve addition of letters, change of letters, or deletion of letters. For this reason it must be hard-coded into the transducer.<br /> <br /> An important next step to the transducer would be to make sure both forms of all verbs are included; this would possibly require consultation with a speaker, as sources do not always include both forms.<br /> <br /> ===Noun Incorporation===<br /> <br /> Hiaki nouns and pronouns can be incorporated into verbs, losing their case morphology and becoming part of the single word. The transducer is structured to handle incorporation of arbitrary nouns into arbitrary verbs. <br /> <br /> Due to the computational complexity of noun incorporation, it is currently disabled. However, pronoun incorporation, which occurs vastly more frequently in the corpus, remains enabled.<br /> <br /> ===Reduplication===<br /> <br /> Many Hiaki verbs reduplicate productively to indicate, for instance, the habitual aspect on verbs, or plurality on adjectives. Functionality for reduplication is partially implemented using .twol rules. However, due to the complexity of Hiaki reduplication and the idiosyncracies across verbs regarding what part of the verb reduplicates, the implementation is not yet complete; in particular, it can only handle reduplication of single open or closed syllables, which must not begin with a vowel. (For example, the transducer can correctly analyze vivicha, the reduplicated form of vicha `see&#039;.)<br /> <br /> Some &quot;reduplications&quot; are not actually reduplications; these appear on verbs which have a different morphological change in place of reduplication for the same semantic effect. Such &quot;reduplications&quot;, which can be, for example, gemination of a consonant in the middle of the word, are hard-coded.<br /> <br /> Examples of Reduplication Types:<br /> *wiuta to wiwiuta (open syllable reduplication)<br /> *kupikte to kupikupikte (two open syllable reduplication)<br /> *hakta to hakhakta (closed syllable reduplication)<br /> *maveta to mavveta (gemination -- not actually reduplication)<br /> <br /> ===Spellrelax===<br /> <br /> A copy of the spellrelax file originally from [here http://svn.code.sf.net/p/apertium/svn/incubator/apertium-wal/dev/spellrelax.twol] is used to relax apostrophe-like symbols into a single symbol, as such symbols can appear inside Hiaki words.<br /> <br /> ==Next Steps==<br /> <br /> Several major grammatical aspects of the Hiaki language remain unimplemented, for example:<br /> <br /> ===Number Suppletion===<br /> <br /> Many Hiaki verbs completely change root form depending on the number of one of their arguments, usually the object if the verb is transitive and the subject otherwise. This is implemented in the transducer with &lt;sg&gt; and &lt;pl&gt; tags on verbs for those verbs known to exhibit number agreement; however, it is often the case in certain word databases that it is not made clear if a verb root is singular, plural, or both. More work needs to be done with the transducer to implement this properly.<br /> <br /> Also, currently the transducer has different lemmas for the two verbs, the singular and the plural form. To create any effective machine translation system in the future, it is likely these lemmas would need to be collapsed into one, which would require a small restructure of the transducer. For the purposes of the current transducer, the lemmas are kept distinct to avoid the confusion of a lemma with no resemblance to its form; for certain uses of the transducer, such as automating glossing, this is likely better. <br /> <br /> ===Vowel Change Rules===<br /> <br /> The vowel change rules in Hiaki are fairly complex, and the forms of affixes often vary in slight ways depending on the root. The transducer needs to be improved by first identifying the regularities and patterns in the alternation of affixes, then implementing them, probably through .twol. Currently some such patterns are implemented, but normally the transducer overgenerates, placing every form of a given affix on any root. There are also likely forms we have not encountered to put in the transducer.<br /> <br /> ===Root Change Rules===<br /> <br /> Many roots in Hiaki change form due to a variety of reasons, from number suppletion to the presence of a derivational suffix to being incorporated with another word to apparently idiosyncratic changes based on specific suffixes, such as wohoʼoria-po and wohoʼok-u, both of which mean `in a hole&#039; but which have slightly different roots in addition to different locative suffixes. Before the transducer can be improved to understand these changes, it must be better understood what exactly conditions them. This requires research into the work done on Hiaki and possibly consultation with speakers.<br /> <br /> ===Dealing with Spanish===<br /> <br /> A large amount of text available in Hiaki comes from transcribed interviews with native speakers, who often switch briefly into Spanish while speaking Hiaki. Several very common Spanish words are included in the transducer as loanwords, but there may be a better way to deal with switching into Spanish. This should be considered going forward in the transducer&#039;s development.<br /> <br /> ==Evaluation==<br /> <br /> Coverage was tested over a corpus (corpus/yaq.corpus.txt). Little text is readily available in Hiaki, so the corpus used only has 3254 words. To continue work on the transducer, it is likely a larger corpus would need to be compiled. <br /> <br /> Currently, the transducer tokenizes 4018 words in the corpus (including punctuation), and has coverage 66.48%.<br /> <br /> Precision and recall were evaluated on a small segment of 88 words of the corpus. The output of the transducer at the time of testing is stored, in CG format, in corpus/yaq.corpus.cg.txt; the correctly annotated version is stored in corpus/yaq.corpus.annotated.txt. Precision over this very small section is 80.53097%, and recall is 91.00000%. These high values are achieved because this is a well-understood section of the corpus (why it was chosen to be annotated), and thus the transducer is over-fit to it, having been built in reference to this part of the corpus.<br /> <br /> The transducer was then modified to improve precision and recall on yaq.corpus.cg.txt. The new output, in CG format, is storied in corpus/yaq.corpus.cg.new.txt. A new annotated version was made as well at corpus/yaq.corpus.cg.annotated.txt, as the transducer was also updated to parse quotation marks, which are now included in the new CG formatted documents. The new annotated corpus only differs from the old one in that it includes quotation marks. After the modification, precision was 77.41935% and recall was 92.30769%. The reason precision decreased may have to do with certain forms, which previously were not analyzed at all, now receiving multiple possibly correct analyses.<br /> <br /> [[Category:sp17_FinalProjects]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=User:Jmundo1/Final_project&diff=5364 2017-05-10T23:45:22Z

<p>Jmundo1: </p> <hr /> <div>=Transducer for Hiaki=<br /> <br /> For my final project, I created a prototype transducer for the Hiaki language, a language extremely lacking in computational resources. It is available at [https://github.swarthmore.edu/jmundo1/ling073-yaq Swarthmore&#039;s github].<br /> <br /> ==Implementation==<br /> <br /> The implementation includes the following notable or unusual functionality, in addition to standard treatment of nouns, verbs, pronouns, postpositions, adjectives, adverbs, and other common Hiaki word types:<br /> <br /> ===Variation in Verb Forms===<br /> <br /> Hiaki verb forms exhibit somewhat idiosyncratic variation in their roots depending on whether derivational suffixation appears. This process could not be automated because of the idiosyncrasy, but the lexicon is built so both forms of a root verb can be placed and the transducer will be sure only to generate the right one.<br /> <br /> Examples:<br /> * poona (without morphology) to pon (with morphology)<br /> * kiima (without morphology) to kima&#039;a (with morphology)<br /> * ne&#039;e (without morphology) to ni&#039;i (with morphology)<br /> <br /> As the above examples illustrate, this change is unpredictable; it can involve addition of letters, change of letters, or deletion of letters. For this reason it must be hard-coded into the transducer.<br /> <br /> An important next step to the transducer would be to make sure both forms of all verbs are included; this would possibly require consultation with a speaker, as sources do not always include both forms.<br /> <br /> ===Noun Incorporation===<br /> <br /> Hiaki nouns and pronouns can be incorporated into verbs, losing their case morphology and becoming part of the single word. The transducer is structured to handle incorporation of arbitrary nouns into arbitrary verbs. <br /> <br /> Due to the computational complexity of noun incorporation, it is currently disabled. However, pronoun incorporation, which occurs vastly more frequently in the corpus, remains enabled.<br /> <br /> ===Reduplication===<br /> <br /> Many Hiaki verbs reduplicate productively to indicate, for instance, the habitual aspect on verbs, or plurality on adjectives. Functionality for reduplication is partially implemented using .twol rules. However, due to the complexity of Hiaki reduplication and the idiosyncracies across verbs regarding what part of the verb reduplicates, the implementation is not yet complete; in particular, it can only handle reduplication of single open or closed syllables, which must not begin with a vowel. (For example, the transducer can correctly analyze vivicha, the reduplicated form of vicha `see&#039;.)<br /> <br /> Some &quot;reduplications&quot; are not actually reduplications; these appear on verbs which have a different morphological change in place of reduplication for the same semantic effect. Such &quot;reduplications&quot;, which can be, for example, gemination of a consonant in the middle of the word, are hard-coded.<br /> <br /> Examples of Reduplication Types:<br /> *wiuta to wiwiuta (open syllable reduplication)<br /> *kupikte to kupikupikte (two open syllable reduplication)<br /> *hakta to hakhakta (closed syllable reduplication)<br /> *maveta to mavveta (gemination -- not actually reduplication)<br /> <br /> ===Spellrelax===<br /> <br /> A copy of the spellrelax file originally from [here http://svn.code.sf.net/p/apertium/svn/incubator/apertium-wal/dev/spellrelax.twol] is used to relax apostrophe-like symbols into a single symbol, as such symbols can appear inside Hiaki words.<br /> <br /> ==Next Steps==<br /> <br /> Several major grammatical aspects of the Hiaki language remain unimplemented, for example:<br /> <br /> ===Number Suppletion===<br /> <br /> Many Hiaki verbs completely change root form depending on the number of one of their arguments, usually the object if the verb is transitive and the subject otherwise. This is implemented in the transducer with &lt;sg&gt; and &lt;pl&gt; tags on verbs for those verbs known to exhibit number agreement; however, it is often the case in certain word databases that it is not made clear if a verb root is singular, plural, or both. More work needs to be done with the transducer to implement this properly.<br /> <br /> Also, currently the transducer has different lemmas for the two verbs, the singular and the plural form. To create any effective machine translation system in the future, it is likely these lemmas would need to be collapsed into one, which would require a small restructure of the transducer. For the purposes of the current transducer, the lemmas are kept distinct to avoid the confusion of a lemma with no resemblance to its form; for certain uses of the transducer, such as automating glossing, this is likely better. <br /> <br /> ===Vowel Change Rules===<br /> <br /> The vowel change rules in Hiaki are fairly complex, and the forms of affixes often vary in slight ways depending on the root. The transducer needs to be improved by first identifying the regularities and patterns in the alternation of affixes, then implementing them, probably through .twol. Currently some such patterns are implemented, but normally the transducer overgenerates, placing every form of a given affix on any root. There are also likely forms we have not encountered to put in the transducer.<br /> <br /> ===Dealing with Spanish===<br /> <br /> A large amount of text available in Hiaki comes from transcribed interviews with native speakers, who often switch briefly into Spanish while speaking Hiaki. Several very common Spanish words are included in the transducer as loanwords, but there may be a better way to deal with switching into Spanish. This should be considered going forward in the transducer&#039;s development.<br /> <br /> ==Evaluation==<br /> <br /> Coverage was tested over a corpus (corpus/yaq.corpus.txt). Little text is readily available in Hiaki, so the corpus used only has 3254 words. To continue work on the transducer, it is likely a larger corpus would need to be compiled. <br /> <br /> Currently, the transducer tokenizes 4018 words in the corpus (including punctuation), and has coverage 66.48%.<br /> <br /> Precision and recall were evaluated on a small segment of 88 words of the corpus. The output of the transducer at the time of testing is stored, in cg format, in corpus/yaq.corpus.cg.txt; the correctly annotated version is stored in corpus/yaq.corpus.annotated.txt. Precision over this very small section is 80.53097%, and recall is 91.00000%. These high values are achieved because this is a well-understood section of the corpus (why it was chosen to be annotated), and thus the transducer is over-fit to it, having been built in reference to this part of the corpus.<br /> <br /> [[Category:sp17_FinalProjects]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=User:Jmundo1/Final_project&diff=5362 2017-05-10T21:51:13Z

<p>Jmundo1: </p> <hr /> <div>=Transducer for Hiaki=<br /> <br /> For my final project, I created a prototype transducer for the Hiaki language, a language extremely lacking in computational resources. It is available at [https://github.swarthmore.edu/jmundo1/ling073-yaq Swarthmore&#039;s github].<br /> <br /> ==Implementation==<br /> <br /> The implementation includes the following notable or unusual functionality, in addition to standard treatment of nouns, verbs, pronouns, postpositions, adjectives, adverbs, and other common Hiaki word types:<br /> <br /> ===Variation in Verb Forms===<br /> <br /> Hiaki verb forms exhibit somewhat idiosyncratic variation in their roots depending on whether derivational suffixation appears. This process could not be automated because of the idiosyncrasy, but the lexicon is built so both forms of a root verb can be placed and the transducer will be sure only to generate the right one.<br /> <br /> Examples:<br /> * poona (without morphology) to pon (with morphology)<br /> * kiima (without morphology) to kima&#039;a (with morphology)<br /> * ne&#039;e (without morphology) to ni&#039;i (with morphology)<br /> <br /> As the above examples illustrate, this change is unpredictable; it can involve addition of letters, change of letters, or deletion of letters. For this reason it must be hard-coded into the transducer.<br /> <br /> An important next step to the transducer would be to make sure both forms of all verbs are included; this would possibly require consultation with a speaker, as sources do not always include both forms.<br /> <br /> ===Noun Incorporation===<br /> <br /> Hiaki nouns and pronouns can be incorporated into verbs, losing their case morphology and becoming part of the single word. The transducer is structured to handle incorporation of arbitrary nouns into arbitrary verbs. <br /> <br /> Due to the computational complexity of noun incorporation, it is currently disabled. However, pronoun incorporation, which occurs vastly more frequently in the corpus, remains enabled.<br /> <br /> ===Reduplication===<br /> <br /> Many Hiaki verbs reduplicate productively to indicate, for instance, the habitual aspect on verbs, or plurality on adjectives. Functionality for reduplication is partially implemented using .twol rules. However, due to the complexity of Hiaki reduplication and the idiosyncracies across verbs regarding what part of the verb reduplicates, the implementation is not yet complete; in particular, it can only handle reduplication of single open or closed syllables, which must not begin with a vowel. (For example, the transducer can correctly analyze vivicha, the reduplicated form of vicha `see&#039;.)<br /> <br /> Some &quot;reduplications&quot; are not actually reduplications; these appear on verbs which have a different morphological change in place of reduplication for the same semantic effect. Such &quot;reduplications&quot;, which can be, for example, gemination of a consonant in the middle of the word, are hard-coded.<br /> <br /> Examples of Reduplication Types:<br /> *wiuta to wiwiuta (open syllable reduplication)<br /> *kupikte to kupikupikte (two open syllable reduplication)<br /> *hakta to hakhakta (closed syllable reduplication)<br /> *maveta to mavveta (gemination -- not actually reduplication)<br /> <br /> ===Spellrelax===<br /> <br /> A copy of the spellrelax file originally from [here http://svn.code.sf.net/p/apertium/svn/incubator/apertium-wal/dev/spellrelax.twol] is used to relax apostrophe-like symbols into a single symbol, as such symbols can appear inside Hiaki words.<br /> <br /> ==Next Steps==<br /> <br /> Several major grammatical aspects of the Hiaki language remain unimplemented, for example:<br /> <br /> ===Number Suppletion===<br /> <br /> Many Hiaki verbs completely change root form depending on the number of one of their arguments, usually the object if the verb is transitive and the subject otherwise. This is implemented in the transducer with &lt;sg&gt; and &lt;pl&gt; tags on verbs for those verbs known to exhibit number agreement; however, it is often the case in certain word databases that it is not made clear if a verb root is singular, plural, or both. More work needs to be done with the transducer to implement this properly.<br /> <br /> Also, currently the transducer has different lemmas for the two verbs, the singular and the plural form. To create any effective machine translation system in the future, it is likely these lemmas would need to be collapsed into one, which would require a small restructure of the transducer. For the purposes of the current transducer, the lemmas are kept distinct to avoid the confusion of a lemma with no resemblance to its form; for certain uses of the transducer, such as automating glossing, this is likely better. <br /> <br /> ===Vowel Change Rules===<br /> <br /> The vowel change rules in Hiaki are fairly complex, and the forms of affixes often vary in slight ways depending on the root. The transducer needs to be improved by first identifying the regularities and patterns in the alternation of affixes, then implementing them, probably through .twol. Currently some such patterns are implemented, but normally the transducer overgenerates, placing every form of a given affix on any root. There are also likely forms we have not encountered to put in the transducer.<br /> <br /> ===Dealing with Spanish===<br /> <br /> A large amount of text available in Hiaki comes from transcribed interviews with native speakers, who often switch briefly into Spanish while speaking Hiaki. Several very common Spanish words are included in the transducer as loanwords, but there may be a better way to deal with switching into Spanish. This should be considered going forward in the transducer&#039;s development.<br /> <br /> ==Evaluation==<br /> <br /> Coverage was tested over a corpus. Little text is readily available in Hiaki, so the corpus used only has 3254 words. To continue work on the transducer, it is likely a larger corpus would need to be compiled.<br /> <br /> Currently, the transducer tokenizes 4018 words in the corpus (including punctuation), and has coverage 66.48%.<br /> <br /> <br /> <br /> [[Category:sp17_FinalProjects]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=User:Jmundo1/Final_project&diff=5361 2017-05-10T21:48:46Z

<p>Jmundo1: </p> <hr /> <div>=Transducer for Hiaki=<br /> <br /> For my final project, I created a prototype transducer for the Hiaki language, a language extremely lacking in computational resources. It is available at [https://github.swarthmore.edu/jmundo1/ling073-yaq Swarthmore&#039;s github].<br /> <br /> ==Implementation==<br /> <br /> The implementation includes the following notable or unusual functonality, in addition to standard treatment of nouns, verbs, pronouns, postpositions, adjectives, adverbs, and other common Hiaki word types:<br /> <br /> ===Variation in Verb Forms===<br /> <br /> Hiaki verb forms exhibit somewhat idiosyncratic variation in their roots depending on whether derivational suffixation appears. This process could not be automated because of the idiosyncracy, but the lexicon is built so both forms of a root verb can be placed and the transducer will be sure only to generate the right one.<br /> <br /> Examples:<br /> * poona (without morphology) to pon (with morphology)<br /> * kiima (without morphology) to kima&#039;a (with morphology)<br /> * ne&#039;e (without morphology) to ni&#039;i (with morphology)<br /> <br /> As the above examples illustrate, this change is unpredictable; it can involve addition of letters, change of letters, or deletion of letters. For this reason it must be hard-coded into the transducer.<br /> <br /> An important next step to the transducer would be to make sure both forms of all verbs are included; this would possibly require consultation with a speaker, as sources do not always include both forms.<br /> <br /> ===Noun Incorporation===<br /> <br /> Hiaki nouns and pronouns can be incorporated into verbs, losing their case morphology and becoming part of the single word. The transducer is structured to handle incorporation of arbitrary nouns into arbitrary verbs. <br /> <br /> Due to the computational complexity of noun incorporaiton, it is currently disabled. However, pronoun incorporation, which occurs vastly more frequently in the corpus, remains enabled.<br /> <br /> ===Reduplication===<br /> <br /> Many Hiaki verbs reduplicate productively to indicate, for instance, the habitual aspect on verbs, or plurality on adjectives. Functionality for reduplication is partially implemented using .twol rules. However, due to the complexity of Hiaki reduplication and the idiosyncracies across verbs regarding what part of the verb reduplicates, the implementation is not yet complete; in particular, it can only handle reduplication of single open or closed syllables, which must not begin with a vowel. (For example, the transducer can correctly analyze vivicha, the reduplicated form of vicha `see&#039;.)<br /> <br /> Some &quot;reduplications&quot; are not actually reduplications; these appear on verbs which have a different morphological change in place of reduplication for the same semantic effect. Such &quot;reduplications&quot;, which can be, for example, gemination of a consonant in the middle of the word, are hard-coded.<br /> <br /> Examples of Reduplication Types:<br /> *wiuta to wiwiuta (open syllable reduplication)<br /> *kupikte to kupikupikte (two open syllable reduplication)<br /> *hakta to hakhakta (closed syllable reduplication)<br /> *maveta to mavveta (gemination -- not actually reduplication)<br /> <br /> ===Spellrelax===<br /> <br /> A copy of the spellrelax file originally from [here http://svn.code.sf.net/p/apertium/svn/incubator/apertium-wal/dev/spellrelax.twol] is used to relax apostrophe-like symbols into a single symbol, as such symbols can appear inside Hiaki words.<br /> <br /> ==Next Steps==<br /> <br /> Several major grammatical aspects of the Hiaki language remain unimplemented, for example:<br /> <br /> ===Number Suppletion===<br /> <br /> Many Hiaki verbs completely change root form depending on the number of one of their arguments, usually the object if the verb is transitive and the subject otherwise. This is implemented in the transducer with &lt;sg&gt; and &lt;pl&gt; tags on verbs for those verbs known to exhibit number agreement; however, it is often the case in certain word databases that it is not made clear if a verb root is singular, plural, or both. More work needs to be done with the transducer to implement this properly.<br /> <br /> Also, currently the transducer has different lemmas for the two verbs, the singular and the plural form. To create any effective machine translation system in the future, it is likely these lemmas would need to be collapsed into one, which would require a small restructure of the transducer. For the purposes of the current transducer, the lemmas are kept distinct to avoid the confusion of a lemma with no resemblance to its form; for certain uses of the transducer, such as automating glossing, this is likely better. <br /> <br /> ===Vowel Change Rules===<br /> <br /> The vowel change rules in Hiaki are fairly complex, and the forms of affixes often vary in slight ways depending on the root. The transducer needs to be improved by first identifying the regularities and patterns in the alternation of affixes, then implementing them, probably through .twol. Currently some such patterns are implemented, but normally the transducer overgenerates, placing every form of a given affix on any root. There are also likely forms we have not encountered to put in the transducer.<br /> <br /> ==Evaluation==<br /> <br /> Coverage was tested over a corpus. Little text is readily available in Hiaki, so the corpus used only has 3254 words. To continue work on the transducer, it is likely a larger corpus would need to be compiled.<br /> <br /> Currently, the transducer tokenizes 4018 words in the corpus (including punctuation), and has coverage 66.48%.<br /> <br /> <br /> <br /> [[Category:sp17_FinalProjects]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=User:Jmundo1/Final_project&diff=5360 2017-05-10T21:45:36Z

<p>Jmundo1: </p> <hr /> <div>=Transducer for Hiaki=<br /> <br /> For my final project, I created a prototype transducer for the Hiaki language, a language extremely lacking in computational resources. It is available at [https://github.swarthmore.edu/jmundo1/ling073-yaq Swarthmore&#039;s github].<br /> <br /> ==Implementation==<br /> <br /> The implementation includes the following notable or unusual functonality, in addition to standard treatment of nouns, verbs, pronouns, postpositions, adjectives, adverbs, and other common Hiaki word types:<br /> <br /> ===Variation in Verb Forms===<br /> <br /> Hiaki verb forms exhibit somewhat idiosyncratic variation in their roots depending on whether derivational suffixation appears. This process could not be automated because of the idiosyncracy, but the lexicon is built so both forms of a root verb can be placed and the transducer will be sure only to generate the right one.<br /> <br /> Examples:<br /> * poona (without morphology) to pon (with morphology)<br /> * kiima (without morphology) to kima&#039;a (with morphology)<br /> * ne&#039;e (without morphology) to ni&#039;i (with morphology)<br /> <br /> As the above examples illustrate, this change is unpredictable; it can involve addition of letters, change of letters, or deletion of letters. For this reason it must be hard-coded into the transducer.<br /> <br /> An important next step to the transducer would be to make sure both forms of all verbs are included; this would possibly require consultation with a speaker, as sources do not always include both forms.<br /> <br /> ===Noun Incorporation===<br /> <br /> Hiaki nouns and pronouns can be incorporated into verbs, losing their case morphology and becoming part of the single word. The transducer is structured to handle incorporation of arbitrary nouns into arbitrary verbs. <br /> <br /> Due to the computational complexity of noun incorporaiton, it is currently disabled. However, pronoun incorporation, which occurs vastly more frequently in the corpus, remains enabled.<br /> <br /> ===Reduplication===<br /> <br /> Many Hiaki verbs reduplicate productively to indicate, for instance, the habitual aspect on verbs, or plurality on adjectives. Functionality for reduplication is partially implemented using .twol rules. However, due to the complexity of Hiaki reduplication and the idiosyncracies across verbs regarding what part of the verb reduplicates, the implementation is not yet complete; in particular, it can only handle reduplication of single open or closed syllables, which must not begin with a vowel. (For example, the transducer can correctly analyze vivicha, the reduplicated form of vicha `see&#039;.)<br /> <br /> Some &quot;reduplications&quot; are not actually reduplications; these appear on verbs which have a different morphological change in place of reduplication for the same semantic effect. Such &quot;reduplications&quot;, which can be, for example, gemination of a consonant in the middle of the word, are hard-coded.<br /> <br /> Examples of Reduplication Types:<br /> *wiuta to wiwiuta (open syllable reduplication)<br /> *kupikte to kupikupikte (two open syllable reduplication)<br /> *hakta to hakhakta (closed syllable reduplication)<br /> *maveta to mavveta (gemination -- not actually reduplication)<br /> <br /> ==Next Steps==<br /> <br /> Several major grammatical aspects of the Hiaki language remain unimplemented, for example:<br /> <br /> ===Number Suppletion===<br /> <br /> Many Hiaki verbs completely change root form depending on the number of one of their arguments, usually the object if the verb is transitive and the subject otherwise. This is implemented in the transducer with &lt;sg&gt; and &lt;pl&gt; tags on verbs for those verbs known to exhibit number agreement; however, it is often the case in certain word databases that it is not made clear if a verb root is singular, plural, or both. More work needs to be done with the transducer to implement this properly.<br /> <br /> Also, currently the transducer has different lemmas for the two verbs, the singular and the plural form. To create any effective machine translation system in the future, it is likely these lemmas would need to be collapsed into one, which would require a small restructure of the transducer. For the purposes of the current transducer, the lemmas are kept distinct to avoid the confusion of a lemma with no resemblance to its form; for certain uses of the transducer, such as automating glossing, this is likely better. <br /> <br /> ===Vowel Change Rules===<br /> <br /> The vowel change rules in Hiaki are fairly complex, and the forms of affixes often vary in slight ways depending on the root. The transducer needs to be improved by first identifying the regularities and patterns in the alternation of affixes, then implementing them, probably through .twol. Currently some such patterns are implemented, but normally the transducer overgenerates, placing every form of a given affix on any root. There are also likely forms we have not encountered to put in the transducer.<br /> <br /> ==Evaluation==<br /> <br /> Coverage was tested over a corpus. Little text is readily available in Hiaki, so the corpus used only has 3254 words. To continue work on the transducer, it is likely a larger corpus would need to be compiled.<br /> <br /> Currently, the transducer tokenizes 4018 words in the corpus (including punctuation), and has coverage 66.48%.<br /> <br /> <br /> <br /> [[Category:sp17_FinalProjects]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=User:Jmundo1/Final_project&diff=5356 2017-05-10T21:29:35Z

<p>Jmundo1: </p> <hr /> <div>=Transducer for Hiaki=<br /> <br /> For my final project, I created a prototype transducer for the Hiaki language, a language extremely lacking in computational resources. It is available at [https://github.swarthmore.edu/jmundo1/ling073-yaq Swarthmore&#039;s github].<br /> <br /> ==Implementation==<br /> <br /> The implementation includes the following notable or unusual functonality, in addition to standard treatment of nouns, verbs, pronouns, postpositions, adjectives, adverbs, and other common Hiaki word types:<br /> <br /> ===Variation in Verb Forms===<br /> <br /> Hiaki verb forms exhibit somewhat idiosyncratic variation in their roots depending on whether derivational suffixation appears. This process could not be automated because of the idiosyncracy, but the lexicon is built so both forms of a root verb can be placed and the transducer will be sure only to generate the right one.<br /> <br /> Examples:<br /> * poona (without morphology) to pon (with morphology)<br /> * kiima (without morphology) to kima&#039;a (with morphology)<br /> * ne&#039;e (without morphology) to ni&#039;i (with morphology)<br /> <br /> As the above examples illustrate, this change is unpredictable; it can involve addition of letters, change of letters, or deletion of letters. For this reason it must be hard-coded into the transducer.<br /> <br /> An important next step to the transducer would be to make sure both forms of all verbs are included; this would possibly require consultation with a speaker, as sources do not always include both forms.<br /> <br /> ===Noun Incorporation===<br /> <br /> Hiaki nouns and pronouns can be incorporated into verbs, losing their case morphology and becoming part of the single word. The transducer is structured to handle incorporation of arbitrary nouns into arbitrary verbs. <br /> <br /> Due to the computational complexity of noun incorporaiton, it is currently disabled. However, pronoun incorporation, which occurs vastly more frequently in the corpus, remains enabled.<br /> <br /> ===Reduplication===<br /> <br /> Many Hiaki verbs reduplicate productively to indicate, for instance, the habitual aspect on verbs, or plurality on adjectives. Functionality for reduplication is partially implemented using .twol rules. However, due to the complexity of Hiaki reduplication and the idiosyncracies across verbs regarding what part of the verb reduplicates, the implementation is not yet complete; in particular, it can only handle reduplication of single open or closed syllables, which must not begin with a vowel. (For example, the transducer can correctly analyze vivicha, the reduplicated form of vicha `see&#039;.)<br /> <br /> Some &quot;reduplications&quot; are not actually reduplications; these appear on verbs which have a different morphological change in place of reduplication for the same semantic effect. Such &quot;reduplications&quot;, which can be, for example, gemination of a consonant in the middle of the word, are hard-coded.<br /> <br /> Examples of Reduplication Types:<br /> *wiuta to wiwiuta (open syllable reduplication)<br /> *kupikte to kupikupikte (two open syllable reduplication)<br /> *hakta to hakhakta (closed syllable reduplication)<br /> *maveta to mavveta (gemination -- not actually reduplication)<br /> <br /> ==Next Steps==<br /> <br /> Several major grammatical aspects of the Hiaki language remain unimplemented, for example:<br /> <br /> ===Number Suppletion===<br /> <br /> Many Hiaki verbs completely change root form depending on the number of one of their arguments, usually the object if the verb is transitive and the subject otherwise. This is implemented in the transducer with &lt;sg&gt; and &lt;pl&gt; tags on verbs for those verbs known to exhibit number agreement; however, it is often the case in certain word databases that it is not made clear if a verb root is singular, plural, or both. More work needs to be done with the transducer to implement this properly.<br /> <br /> Also, currently the transducer has different lemmas for the two verbs, the singular and the plural form. To create any effective machine translation system in the future, it is likely these lemmas would need to be collapsed into one, which would require a small restructure of the transducer. For the purposes of the current transducer, the lemmas are kept distinct to avoid the confusion of a lemma with no resemblance to its form; for certain uses of the transducer, such as automating glossing, this is likely better. <br /> <br /> ===Vowel Change Rules===<br /> <br /> The vowel change rules in Hiaki are fairly complex, and the forms of affixes often vary in slight ways depending on the root. The transducer needs to be improved by first identifying the regularities and patterns in the alternation of affixes, then implementing them, probably through .twol. Currently some such patterns are implemented, but normally the transducer overgenerates, placing every form of a given affix on any root. There are also likely forms we have not encountered to put in the transducer.<br /> <br /> ==Evaluation==<br /> <br /> Coverage was tested over a corpus. Little text is readily available in Hiaki, so the corpus used only has 3254 words. To continue work on the transducer, it is likely a larger corpus would need to be compiled.<br /> <br /> [[Category:sp17_FinalProjects]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=User:Jmundo1/Final_project&diff=5355 2017-05-10T21:28:52Z

<p>Jmundo1: </p> <hr /> <div>=Transducer for Hiaki=<br /> <br /> For my final project, I created a prototype transducer for the Hiaki language, a language extremely lacking in computational resources. It is available at [https://github.swarthmore.edu/jmundo1/ling073-yaq Swarthmore&#039;s github].<br /> <br /> ==Implementation==<br /> <br /> The implementation includes the following notable or unusual functonality, in addition to standard treatment of nouns, verbs, pronouns, postpositions, adjectives, adverbs, and other common Hiaki word types:<br /> <br /> ===Variation in Verb Forms===<br /> <br /> Hiaki verb forms exhibit somewhat idiosyncratic variation in their roots depending on whether derivational suffixation appears. This process could not be automated because of the idiosyncracy, but the lexicon is built so both forms of a root verb can be placed and the transducer will be sure only to generate the right one.<br /> <br /> Examples:<br /> * poona (without morphology) to pon (with morphology)<br /> * kiima (without morphology) to kima&#039;a (with morphology)<br /> * ne&#039;e (without morphology) to ni&#039;i (with morphology)<br /> <br /> As the above examples illustrate, this change is unpredictable; it can involve addition of letters, change of letters, or deletion of letters. For this reason it must be hard-coded into the transducer.<br /> <br /> An important next step to the transducer would be to make sure both forms of all verbs are included; this would require consultation with a speaker, as no dictionaries I have found consistently include both forms.<br /> <br /> ===Noun Incorporation===<br /> <br /> Hiaki nouns and pronouns can be incorporated into verbs, losing their case morphology and becoming part of the single word. The transducer is structured to handle incorporation of arbitrary nouns into arbitrary verbs. <br /> <br /> Due to the computational complexity of noun incorporaiton, it is currently disabled. However, pronoun incorporation, which occurs vastly more frequently in the corpus, remains enabled.<br /> <br /> ===Reduplication===<br /> <br /> Many Hiaki verbs reduplicate productively to indicate, for instance, the habitual aspect on verbs, or plurality on adjectives. Functionality for reduplication is partially implemented using .twol rules. However, due to the complexity of Hiaki reduplication and the idiosyncracies across verbs regarding what part of the verb reduplicates, the implementation is not yet complete; in particular, it can only handle reduplication of single open or closed syllables, which must not begin with a vowel. (For example, the transducer can correctly analyze vivicha, the reduplicated form of vicha `see&#039;.)<br /> <br /> Some &quot;reduplications&quot; are not actually reduplications; these appear on verbs which have a different morphological change in place of reduplication for the same semantic effect. Such &quot;reduplications&quot;, which can be, for example, gemination of a consonant in the middle of the word, are hard-coded.<br /> <br /> Examples of Reduplication Types:<br /> *wiuta to wiwiuta (open syllable reduplication)<br /> *kupikte to kupikupikte (two open syllable reduplication)<br /> *hakta to hakhakta (closed syllable reduplication)<br /> *maveta to mavveta (gemination -- not actually reduplication)<br /> <br /> ==Next Steps==<br /> <br /> Several major grammatical aspects of the Hiaki language remain unimplemented, for example:<br /> <br /> ===Number Suppletion===<br /> <br /> Many Hiaki verbs completely change root form depending on the number of one of their arguments, usually the object if the verb is transitive and the subject otherwise. This is implemented in the transducer with &lt;sg&gt; and &lt;pl&gt; tags on verbs for those verbs known to exhibit number agreement; however, it is often the case in dictionaries and word databases that it is not made clear if a verb root is singular, plural, or both. More work needs to be done with the transducer to implement this properly.<br /> <br /> Also, currently the transducer has different lemmas for the two verbs, the singular and the plural form. To create any effective machine translation system in the future, it is likely these lemmas would need to be collapsed into one, which would require a small restructure of the transducer. For the purposes of the current transducer, the lemmas are kept distinct to avoid the confusion of a lemma with no resemblance to its form; for certain uses of the transducer, such as automating glossing, this is likely better. <br /> <br /> ===Vowel Change Rules===<br /> <br /> The vowel change rules in Hiaki are fairly complex, and the forms of affixes often vary in slight ways depending on the root. The transducer needs to be improved by first identifying the regularities and patterns in the alternation of affixes, then implementing them, probably through .twol. Currently some such patterns are implemented, but normally the transducer overgenerates, placing every form of a given affix on any root. There are also likely forms we have not encountered to put in the transducer.<br /> <br /> ==Evaluation==<br /> <br /> Coverage was tested over a corpus. Little text is readily available in Hiaki, so the corpus used only has 3254 words. To continue work on the transducer, it is likely a larger corpus would need to be compiled.<br /> <br /> [[Category:sp17_FinalProjects]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=User:Jmundo1/Final_project&diff=5350 2017-05-10T21:14:12Z

<p>Jmundo1: </p> <hr /> <div>=Transducer for Hiaki=<br /> <br /> For my final project, I created a prototype transducer for the Hiaki language, a language extremely lacking in computational resources. It is available at [https://github.swarthmore.edu/jmundo1/ling073-yaq Swarthmore&#039;s github].<br /> <br /> ==Implementation==<br /> <br /> The implementation includes the following notable or unusual functonality, in addition to standard treatment of nouns, verbs, pronouns, postpositions, adjectives, adverbs, and other common Hiaki word types:<br /> <br /> ===Variation in Verb Forms===<br /> <br /> Hiaki verb forms exhibit somewhat idiosyncratic variation in their roots depending on whether derivational suffixation appears. This process could not be automated because of the idiosyncracy, but the lexicon is built so both forms of a root verb can be placed and the transducer will be sure only to generate the right one.<br /> <br /> Examples:<br /> * poona (without morphology) to pon (with morphology)<br /> * kiima (without morphology) to kima&#039;a (with morphology)<br /> * ne&#039;e (without morphology) to ni&#039;i (with morphology)<br /> <br /> As the above examples illustrate, this change is unpredictable; it can involve addition of letters, change of letters, or deletion of letters. For this reason it must be hard-coded into the transducer.<br /> <br /> An important next step to the transducer would be to make sure both forms of all verbs are included; this would require consultation with a speaker, as no dictionaries I have found consistently include both forms.<br /> <br /> ===Noun Incorporation===<br /> <br /> Hiaki nouns and pronouns can be incorporated into verbs, losing their case morphology and becoming part of the single word. The transducer is structured to handle incorporation of arbitrary nouns into arbitrary verbs. <br /> <br /> Due to the computational complexity of noun incorporaiton, it is currently disabled. However, pronoun incorporation, which occurs vastly more frequently in the corpus, remains enabled.<br /> <br /> ===Reduplication===<br /> <br /> Many Hiaki verbs reduplicate productively to indicate, for instance, the habitual aspect on verbs, or plurality on adjectives. Functionality for reduplication is partially implemented using .twol rules. However, due to the complexity of Hiaki reduplication and the idiosyncracies across verbs regarding what part of the verb reduplicates, the implementation is not yet complete; in particular, it can only handle reduplication of single open or closed syllables, which must not begin with a vowel. (For example, the transducer can correctly analyze vivicha, the reduplicated form of vicha `see&#039;.)<br /> <br /> Some &quot;reduplications&quot; are not actually reduplications; these appear on verbs which have a different morphological change in place of reduplication for the same semantic effect. Such &quot;reduplications&quot;, which can be, for example, gemination of a consonant in the middle of the word, are hard-coded.<br /> <br /> Examples of Reduplication Types:<br /> *wiuta to wiwiuta (open syllable reduplication)<br /> *kupikte to kupikupikte (two open syllable reduplication)<br /> *hakta to hakhakta (closed syllable reduplication)<br /> *maveta to mavveta (gemination -- not actually reduplication)<br /> <br /> ==Next Steps==<br /> <br /> Several major grammatical aspects of the Hiaki language remain unimplemented, for example:<br /> <br /> ===Number Suppletion===<br /> <br /> Many Hiaki verbs completely change root form depending on the number of one of their arguments, usually the object if the verb is transitive and the subject otherwise. This is implemented in the transducer with &lt;sg&gt; and &lt;pl&gt; tags on verbs for those verbs known to exhibit number agreement; however, it is often the case in dictionaries and word databases that it is not made clear if a verb root is singular, plural, or both. More work needs to be done with the transducer to implement this properly.<br /> <br /> ===Vowel Change Rules===<br /> <br /> The vowel change rules in Hiaki are fairly complex, and the forms of affixes often vary in slight ways depending on the root. The transducer needs to be improved by first identifying the regularities and patterns in the alternation of affixes, then implementing them, probably through .twol. Currently some such patterns are implemented, but normally the transducer overgenerates, placing every form of a given affix on any root. There are also likely forms we have not encountered to put in the transducer.<br /> <br /> ==Evaluation==<br /> <br /> Coverage was tested over a corpus. Little text is readily available in Hiaki, so the corpus used only has 3254 words. To continue work on the transducer, it is likely a larger corpus would need to be compiled.<br /> <br /> [[Category:sp17_FinalProjects]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=User:Jmundo1/Final_project&diff=5347 2017-05-10T21:05:02Z

<p>Jmundo1: </p> <hr /> <div>=Transducer for Hiaki=<br /> <br /> For my final project, I created a prototype transducer for the Hiaki language, a language extremely lacking in computational resources. It is available at [https://github.swarthmore.edu/jmundo1/ling073-yaq Swarthmore&#039;s github].<br /> <br /> ==Implementation==<br /> <br /> The implementation includes the following notable or unusual functonality, in addition to standard treatment of nouns, verbs, pronouns, postpositions, adjectives, adverbs, and other common Hiaki word types:<br /> <br /> ===Variation in Verb Forms===<br /> <br /> Hiaki verb forms exhibit somewhat idiosyncratic variation in their roots depending on whether derivational suffixation appears. This process could not be automated because of the idiosyncracy, but the lexicon is built so both forms of a root verb can be placed and the transducer will be sure only to generate the right one.<br /> <br /> Examples:<br /> * poona (without morphology) to pon (with morphology)<br /> * kiima (without morphology) to kima&#039;a (with morphology)<br /> * ne&#039;e (without morphology) to ni&#039;i (with morphology)<br /> <br /> As the above examples illustrate, this change is unpredictable; it can involve addition of letters, change of letters, or deletion of letters. For this reason it must be hard-coded into the transducer.<br /> <br /> An important next step to the transducer would be to make sure both forms of all verbs are included; this would require consultation with a speaker, as no dictionaries I have found consistently include both forms.<br /> <br /> ===Noun Incorporation===<br /> <br /> Hiaki nouns and pronouns can be incorporated into verbs, losing their case morphology and becoming part of the single word. The transducer is structured to handle incorporation of arbitrary nouns into arbitrary verbs. <br /> <br /> Due to the computational complexity of noun incorporaiton, it is currently disabled. However, pronoun incorporation, which occurs vastly more frequently in the corpus, remains enabled.<br /> <br /> ===Reduplication===<br /> <br /> Many Hiaki verbs reduplicate productively to indicate, for instance, the habitual aspect on verbs, or plurality on adjectives. Functionality for reduplication is partially implemented using .twol rules. However, due to the complexity of Hiaki reduplication and the idiosyncracies across verbs regarding what part of the verb reduplicates, the implementation is not yet complete; in particular, it can only handle reduplication of single open or closed syllables, which must not begin with a vowel.<br /> <br /> Some &quot;reduplications&quot; are not actually reduplications; these appear on verbs which have a different morphological change in place of reduplication for the same semantic effect. Such &quot;reduplications&quot;, which can be, for example, gemination of a consonant in the middle of the word, are hard-coded.<br /> <br /> Examples of Reduplication Types:<br /> *wiuta to wiwiuta (open syllable reduplication)<br /> *kupikte to kupikupikte (two open syllable reduplication)<br /> *hakta to hakhakta (closed syllable reduplication)<br /> *maveta to mavveta (gemination -- not actually reduplication)<br /> <br /> ==Next Steps==<br /> <br /> Several major grammatical aspects of the Hiaki language remain unimplemented, for example:<br /> <br /> ===Number Suppletion===<br /> <br /> Many Hiaki verbs completely change root form depending on the number of one of their arguments, usually the object if the verb is transitive and the subject otherwise. This is implemented in the transducer with &lt;sg&gt; and &lt;pl&gt; tags on verbs for those verbs known to exhibit number agreement; however, it is often the case in dictionaries and word databases that it is not made clear if a verb root is singular, plural, or both. More work needs to be done with the transducer to implement this properly.<br /> <br /> ===Vowel Change Rules===<br /> <br /> The vowel change rules in Hiaki are fairly complex, and the forms of affixes often vary in slight ways depending on the root. The transducer needs to be improved by first identifying the regularities and patterns in the alternation of affixes, then implementing them, probably through .twol. Currently some such patterns are implemented, but normally the transducer overgenerates, placing every form of a given affix on any root. There are also likely forms we have not encountered to put in the transducer.<br /> <br /> ==Evaluation==<br /> <br /> Coverage was tested over a corpus. Little text is readily available in Hiaki, so the corpus used only has 3254 words. To continue work on the transducer, it is likely a larger corpus would need to be compiled.<br /> <br /> [[Category:sp17_FinalProjects]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=User:Jmundo1/Final_project&diff=5346 2017-05-10T21:01:53Z

<p>Jmundo1: </p> <hr /> <div>=Transducer for Hiaki=<br /> <br /> For my final project, I created a prototype transducer for the Hiaki language, a language extremely lacking in computational resources. It is available at [https://github.swarthmore.edu/jmundo1/ling073-yaq Swarthmore&#039;s github].<br /> <br /> ==Implementation==<br /> <br /> The implementation includes the following notable or unusual functonality, in addition to standard treatment of nouns, verbs, pronouns, postpositions, adjectives, adverbs, and other common Hiaki word types:<br /> <br /> ===Variation in Verb Forms===<br /> <br /> Hiaki verb forms exhibit somewhat idiosyncratic variation in their roots depending on whether derivational suffixation appears. This process could not be automated because of the idiosyncracy, but the lexicon is built so both forms of a root verb can be placed and the transducer will be sure only to generate the right one.<br /> <br /> Examples:<br /> * poona (without morphology) to pon (with morphology)<br /> * kiima (without morphology) to kima&#039;a (with morphology)<br /> * ne&#039;e (without morphology) to ni&#039;i (with morphology)<br /> <br /> As the above examples illustrate, this change is unpredictable; it can involve addition of letters, change of letters, or deletion of letters. For this reason it must be hard-coded into the transducer.<br /> <br /> An important next step to the transducer would be to make sure both forms of all verbs are included; this would require consultation with a speaker, as no dictionaries I have found consistently include both forms.<br /> <br /> ===Noun Incorporation===<br /> <br /> Hiaki nouns and pronouns can be incorporated into verbs, losing their case morphology and becoming part of the single word. The transducer is structured to handle incorporation of arbitrary nouns into arbitrary verbs. <br /> <br /> Due to the computational complexity of noun incorporaiton, it is currently disabled. However, pronoun incorporation, which occurs vastly more frequently in the corpus, remains enabled.<br /> <br /> ===Reduplication===<br /> <br /> Many Hiaki verbs reduplicate productively to indicate, for instance, the habitual aspect on verbs, or plurality on adjectives. Functionality for reduplication is partially implemented using .twol rules. However, due to the complexity of Hiaki reduplication and the idiosyncracies across verbs regarding what part of the verb reduplicates, the implementation should not yet be considered complete.<br /> <br /> Some &quot;reduplications&quot; are not actually reduplications; these appear on verbs which have a different morphological change in place of reduplication for the same semantic effect. Such &quot;reduplications&quot;, which can be, for example, gemination of a consonant in the middle of the word, are hard-coded.<br /> <br /> Examples of Reduplication Types:<br /> *wiuta to wiwiuta (open syllable reduplication)<br /> *kupikte to kupikupikte (two open syllable reduplication)<br /> *hakta to hakhakta (closed syllable reduplication)<br /> *maveta to mavveta (gemination -- not actually reduplication)<br /> <br /> ==Next Steps==<br /> <br /> Several major grammatical aspects of the Hiaki language remain unimplemented, for example:<br /> <br /> ===Number Suppletion===<br /> <br /> Many Hiaki verbs completely change root form depending on the number of one of their arguments, usually the object if the verb is transitive and the subject otherwise. This is implemented in the transducer with &lt;sg&gt; and &lt;pl&gt; tags on verbs for those verbs known to exhibit number agreement; however, it is often the case in dictionaries and word databases that it is not made clear if a verb root is singular, plural, or both. More work needs to be done with the transducer to implement this properly.<br /> <br /> ===Vowel Change Rules===<br /> <br /> The vowel change rules in Hiaki are fairly complex, and the forms of affixes often vary in slight ways depending on the root. The transducer needs to be improved by first identifying the regularities and patterns in the alternation of affixes, then implementing them, probably through .twol. Currently some such patterns are implemented, but normally the transducer overgenerates, placing every form of a given affix on any root. There are also likely forms we have not encountered to put in the transducer.<br /> <br /> ==Evaluation==<br /> <br /> Coverage was tested over a corpus. Little text is readily available in Hiaki, so the corpus used only has 3254 words. To continue work on the transducer, it is likely a larger corpus would need to be compiled.<br /> <br /> [[Category:sp17_FinalProjects]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=User:Jmundo1/Final_project&diff=5345 2017-05-10T21:00:15Z

<p>Jmundo1: </p> <hr /> <div>=Transducer for Hiaki=<br /> <br /> For my final project, I created a prototype transducer for the Hiaki language, a language extremely lacking in computational resources. <br /> <br /> ==Implementation==<br /> <br /> The implementation includes the following notable or unusual functonality, in addition to standard treatment of nouns, verbs, pronouns, postpositions, adjectives, adverbs, and other common Hiaki word types:<br /> <br /> ===Variation in Verb Forms===<br /> <br /> Hiaki verb forms exhibit somewhat idiosyncratic variation in their roots depending on whether derivational suffixation appears. This process could not be automated because of the idiosyncracy, but the lexicon is built so both forms of a root verb can be placed and the transducer will be sure only to generate the right one.<br /> <br /> Examples:<br /> * poona (without morphology) to pon (with morphology)<br /> * kiima (without morphology) to kima&#039;a (with morphology)<br /> * ne&#039;e (without morphology) to ni&#039;i (with morphology)<br /> <br /> As the above examples illustrate, this change is unpredictable; it can involve addition of letters, change of letters, or deletion of letters. For this reason it must be hard-coded into the transducer.<br /> <br /> An important next step to the transducer would be to make sure both forms of all verbs are included; this would require consultation with a speaker, as no dictionaries I have found consistently include both forms.<br /> <br /> ===Noun Incorporation===<br /> <br /> Hiaki nouns and pronouns can be incorporated into verbs, losing their case morphology and becoming part of the single word. The transducer is structured to handle incorporation of arbitrary nouns into arbitrary verbs. <br /> <br /> Due to the computational complexity of noun incorporaiton, it is currently disabled. However, pronoun incorporation, which occurs vastly more frequently in the corpus, remains enabled.<br /> <br /> ===Reduplication===<br /> <br /> Many Hiaki verbs reduplicate productively to indicate, for instance, the habitual aspect on verbs, or plurality on adjectives. Functionality for reduplication is partially implemented using .twol rules. However, due to the complexity of Hiaki reduplication and the idiosyncracies across verbs regarding what part of the verb reduplicates, the implementation should not yet be considered complete.<br /> <br /> Some &quot;reduplications&quot; are not actually reduplications; these appear on verbs which have a different morphological change in place of reduplication for the same semantic effect. Such &quot;reduplications&quot;, which can be, for example, gemination of a consonant in the middle of the word, are hard-coded.<br /> <br /> Examples of Reduplication Types:<br /> *wiuta to wiwiuta (open syllable reduplication)<br /> *kupikte to kupikupikte (two open syllable reduplication)<br /> *hakta to hakhakta (closed syllable reduplication)<br /> *maveta to mavveta (gemination -- not actually reduplication)<br /> <br /> ==Next Steps==<br /> <br /> Several major grammatical aspects of the Hiaki language remain unimplemented, for example:<br /> <br /> ===Number Suppletion===<br /> <br /> Many Hiaki verbs completely change root form depending on the number of one of their arguments, usually the object if the verb is transitive and the subject otherwise. This is implemented in the transducer with &lt;sg&gt; and &lt;pl&gt; tags on verbs for those verbs known to exhibit number agreement; however, it is often the case in dictionaries and word databases that it is not made clear if a verb root is singular, plural, or both. More work needs to be done with the transducer to implement this properly.<br /> <br /> ===Vowel Change Rules===<br /> <br /> The vowel change rules in Hiaki are fairly complex, and the forms of affixes often vary in slight ways depending on the root. The transducer needs to be improved by first identifying the regularities and patterns in the alternation of affixes, then implementing them, probably through .twol. Currently some such patterns are implemented, but normally the transducer overgenerates, placing every form of a given affix on any root. There are also likely forms we have not encountered to put in the transducer.<br /> <br /> ==Evaluation==<br /> <br /> Coverage was tested over a corpus. Little text is readily available in Hiaki, so the corpus used only has 3254 words. To continue work on the transducer, it is likely a larger corpus would need to be compiled.<br /> <br /> [[Category:sp17_FinalProjects]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=User:Jmundo1/Final_project&diff=5344 2017-05-10T20:56:24Z

<p>Jmundo1: </p> <hr /> <div>=Transducer for Hiaki=<br /> <br /> For my final project, I created a prototype transducer for the Hiaki language, a language extremely lacking in computational resources. <br /> <br /> ==Implementation==<br /> <br /> The implementation includes the following notable or unusual functonality, in addition to standard treatment of nouns, verbs, pronouns, postpositions, adjectives, adverbs, and other common Hiaki word types:<br /> <br /> ===Variation in Verb Forms===<br /> <br /> Hiaki verb forms exhibit somewhat idiosyncratic variation in their roots depending on whether derivational suffixation appears. This process could not be automated because of the idiosyncracy, but the lexicon is built so both forms of a root verb can be placed and the transducer will be sure only to generate the right one.<br /> <br /> Examples<br /> * poona (without morphology) to pon (with morphology)<br /> * kiima (without morphology) to kima&#039;a (with morphology)<br /> * ne&#039;e (without morphology) to ni&#039;i (with morphology)<br /> <br /> As the above examples illustrate, this change is unpredictable; it can involve addition of letters, change of letters, or deletion of letters. For this reason it must be hard-coded into the transducer.<br /> <br /> An important next step to the transducer would be to make sure both forms of all verbs are included; this would require consultation with a speaker, as no dictionaries I have found consistently include both forms.<br /> <br /> ===Noun Incorporation===<br /> <br /> Hiaki nouns and pronouns can be incorporated into verbs, losing their case morphology and becoming part of the single word. The transducer is structured to handle incorporation of arbitrary nouns into arbitrary verbs. <br /> <br /> Due to the computational complexity of noun incorporaiton, it is currently disabled. However, pronoun incorporation, which occurs vastly more frequently in the corpus, remains enabled.<br /> <br /> ===Reduplication===<br /> <br /> Many Hiaki verbs reduplicate productively to indicate, for instance, the habitual aspect on verbs, or plurality on adjectives. Functionality for reduplication is partially implemented using .twol rules. However, due to the complexity of Hiaki reduplication and the idiosyncracies across verbs regarding what part of the verb reduplicates, the implementation should not yet be considered complete.<br /> <br /> Some &quot;reduplications&quot; are not actually reduplications; these appear on verbs which have a different morphological change in place of reduplication for the same semantic effect. Such &quot;reduplications&quot;, which can be, for example, gemination of a consonant in the middle of the word, are hard-coded.<br /> <br /> ==Next Steps==<br /> <br /> Several major grammatical aspects of the Hiaki language remain unimplemented, for example:<br /> <br /> ===Number Suppletion===<br /> <br /> Many Hiaki verbs completely change root form depending on the number of one of their arguments, usually the object if the verb is transitive and the subject otherwise. This is implemented in the transducer with &lt;sg&gt; and &lt;pl&gt; tags on verbs for those verbs known to exhibit number agreement; however, it is often the case in dictionaries and word databases that it is not made clear if a verb root is singular, plural, or both. More work needs to be done with the transducer to implement this properly.<br /> <br /> ===Vowel Change Rules===<br /> <br /> The vowel change rules in Hiaki are fairly complex, and the forms of affixes often vary in slight ways depending on the root. The transducer needs to be improved by first identifying the regularities and patterns in the alternation of affixes, then implementing them, probably through .twol. Currently some such patterns are implemented, but normally the transducer overgenerates, placing every form of a given affix on any root. There are also likely forms we have not encountered to put in the transducer.<br /> <br /> ==Evaluation==<br /> <br /> Coverage was tested over a corpus. Little text is readily available in Hiaki, so the corpus used only has 3254 words. To continue work on the transducer, it is likely a larger corpus would need to be compiled.<br /> <br /> [[Category:sp17_FinalProjects]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=User:Jmundo1/Final_project&diff=5339 2017-05-10T20:36:40Z

<p>Jmundo1: </p> <hr /> <div>==Transducer for Hiaki==<br /> <br /> For my final project, I created a prototype transducer for the Hiaki language, a language extremely lacking in computational resources. The implementation includes the following notable or unusual functonality, in addition to standard treatment of nouns, verbs, pronouns, postpositions, adjectives, adverbs, and other common Hiaki word types:<br /> <br /> ===Variation in Verb Forms===<br /> <br /> Hiaki verb forms exhibit somewhat idiosyncratic variation in their roots depending on whether derivational suffixation appears. This process could not be automated because of the idiosyncracy, but the lexicon is built so both forms of a root verb can be placed and the transducer will be sure only to generate the right one.<br /> <br /> ===Noun Incorporation===<br /> <br /> Hiaki nouns and pronouns can be incorporated into verbs, losing their case morphology and becoming part of the single word. The transducer is structured to handle incorporation of arbitrary nouns into arbitrary verbs. <br /> <br /> Due to the computational complexity of noun incorporaiton, it is currently disabled. However, pronoun incorporation, which occurs vastly more frequently in the corpus, remains enabled.<br /> <br /> ===Reduplication===<br /> <br /> Many Hiaki verbs reduplicate productively to indicate, for instance, the habitual aspect on verbs, or plurality on adjectives. Functionality for reduplication is partially implemented using .twol rules. However, due to the complexity of Hiaki reduplication and the idiosyncracies across verbs regarding what part of the verb reduplicates, the implementation should not yet be considered complete.<br /> <br /> ==Evaluation==<br /> <br /> Coverage was tested over a corpus. Little text is readily available in Hiaki, so the corpus used only has 3254 words. To continue work on the transducer, it is likely a larger corpus would need to be compiled.<br /> <br /> [[Category:sp17_FinalProjects]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=User:Jmundo1/Final_project&diff=5338 2017-05-10T20:35:46Z

<p>Jmundo1: Created page with &quot;==Final Project: Transducer for Hiaki== For my final project, I created a prototype transducer for the Hiaki language, a language extremely lacking in computational resources...&quot;</p> <hr /> <div>==Final Project: Transducer for Hiaki==<br /> <br /> For my final project, I created a prototype transducer for the Hiaki language, a language extremely lacking in computational resources. The implementation includes the following notable or unusual functonality, in addition to standard treatment of nouns, verbs, pronouns, postpositions, adjectives, adverbs, and other common Hiaki word types:<br /> <br /> ===Variation in Verb Forms===<br /> <br /> Hiaki verb forms exhibit somewhat idiosyncratic variation in their roots depending on whether derivational suffixation appears. This process could not be automated because of the idiosyncracy, but the lexicon is built so both forms of a root verb can be placed and the transducer will be sure only to generate the right one.<br /> <br /> ===Noun Incorporation===<br /> <br /> Hiaki nouns and pronouns can be incorporated into verbs, losing their case morphology and becoming part of the single word. The transducer is structured to handle incorporation of arbitrary nouns into arbitrary verbs. <br /> <br /> Due to the computational complexity of noun incorporaiton, it is currently disabled. However, pronoun incorporation, which occurs vastly more frequently in the corpus, remains enabled.<br /> <br /> ===Reduplication===<br /> <br /> Many Hiaki verbs reduplicate productively to indicate, for instance, the habitual aspect on verbs, or plurality on adjectives. Functionality for reduplication is partially implemented using .twol rules. However, due to the complexity of Hiaki reduplication and the idiosyncracies across verbs regarding what part of the verb reduplicates, the implementation should not yet be considered complete.<br /> <br /> ==Evaluation==<br /> <br /> Coverage was tested over a corpus. Little text is readily available in Hiaki, so the corpus used only has 3254 words. To continue work on the transducer, it is likely a larger corpus would need to be compiled.<br /> <br /> [[Category:sp17_FinalProjects]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Kikuyu/Universal_Dependencies&diff=4850 2017-04-25T01:35:56Z

<p>Jmundo1: </p> <hr /> <div>This page details Universal Dependencies for Kikuyu, and describes UD annotation and the training of several models on corpora.<br /> <br /> =Evaluation=<br /> <br /> <br /> Here, Corpus 1 is the corpus on which the parsers were trained, and Corpus 2 is another Kikuyu corpus. The withmorph parser was trained on the corpus with morphological information, while the nomorph parser was trained on the corpus without morphological information.<br /> <br /> The following summarizes the sizes of the corpora:<br /> {|class=&quot;wikitable&quot;<br /> |-<br /> |<br /> |Corpus 1<br /> |Corpus 2<br /> |-<br /> |Sentences<br /> |27<br /> |16<br /> |-<br /> |Forms<br /> |197<br /> |110<br /> |}<br /> <br /> The performance of the parsers on the corpora is as follows:<br /> {| class=&quot;wikitable&quot;<br /> |+Evaluation of Parsers on Corpus 1<br /> |-<br /> |<br /> |LAS<br /> |UAS<br /> |-<br /> |withmorph<br /> |60.53%<br /> |63.16%<br /> |-<br /> |nomorph<br /> |49.12%<br /> |57.02%<br /> |}<br /> <br /> For Corpus 1, the withmorph parser fared better on both LAS and UAS metrics, although the degree to which it fared better than the nomorph parser was somewhat higher for the LAS metric.<br /> <br /> {| class=&quot;wikitable&quot;<br /> |+Evaluation of Parsers on Corpus 2<br /> |-<br /> |<br /> |LAS<br /> |UAS<br /> |-<br /> |withmorph<br /> |8.86%<br /> |22.78%<br /> |-<br /> |nomorph<br /> |6.33%<br /> |30.38%<br /> |}<br /> <br /> Both parsers fared much worse on Corpus 2 than on Corpus 1, as expected, given that the parsers were trained on Corpus 1. Surprisingly, the nomorph parser actually had a better UAS score than the withmorph parser for Corpus 2.<br /> <br /> =Dependency Relations=<br /> <br /> In this section, five different dependency relations used in annotating the corpora are detailed.<br /> <br /> ==nmod==<br /> <br /> The nmod relation is, in general, used to mark a nominal modifying another nominal in some way. In Kikuyu, this form of modification appears often; for example, in the noun-associative-noun construction, the second nominal serves as a modifier of the first. There are also numerous constructions in which a noun modifies an adjacent noun. Examples include:<br /> <br /> *The phrase &quot;mbathi thaa inya&quot;, meaning &#039;ten o&#039;clock bus&#039; (literally &#039;bus hour four&#039;), has the noun &quot;thaa&quot; (&#039;hour&#039;) modifying the noun &quot;mbathi&quot; (&#039;bus&#039;) with an nmod relationship.<br /> <br /> *The phrase &quot;marwa ma tax&quot;, meaning &#039;tax letter&#039; (literally &#039;letter ASSOC tax&#039;), has the nominal headed by &quot;tax&quot; (&#039;tax&#039;) modifying the noun &quot;marwa&quot; (&#039;letter&#039;) with an nmod relationship.<br /> <br /> ==case==<br /> <br /> The case relation is, in general, used to mark something like a preposition which acts as a case-assigner for a nominal. In Kikuyu, we have chosen to analyze the relationship between an associative marker and the next word as one of case, as the associative plays a similar role to that of a word like &quot;of&quot; in English. Furthermore, the associative is used in possessive constructions, which usually involve a case relationship. The case relationship also appears with prepositions.<br /> <br /> *The phrase &quot;ndarĩkia na gym&quot;, meaning &#039;I finished with the gym&#039; (literally &#039;I-finished with gym&#039;), has the preposition &quot;na&quot; (&#039;with&#039;) in a case relationship with the nominal &quot;gym&quot; (&#039;gym&#039;).<br /> <br /> *The phrase &quot;Mũthenya Ũmwe wa Wambũi&quot;, meaning &#039;A Day in the Life of Wambũi&quot; (literally &#039;day one ASSOC Wambũi&#039;), has the associative &quot;wa&quot; in a case relationship with the nominal &quot;Wambũi&quot;.<br /> <br /> ==num==<br /> <br /> The num relation is used when a number is modifying a nominal. This construction appears often in Kikuyu, with the number word following the nominal.<br /> <br /> *The phrase &quot;thaa ithatũ kĩroko&quot;, meaning &#039;9 in the morning&#039; (literally &#039;hour three morning&#039;), has the number &quot;ithatũ&quot; modifying the nominal &quot;thaa&quot; (&#039;hour&#039;) with a nummod dependency. This is generally how time is expressed in Kikuyu.<br /> <br /> *The phrase &quot;mĩrongo ĩrĩ&quot;, meaning &#039;twenty&#039; (literally &#039;set-of-ten two&#039;), has the number &quot;ĩrĩ&quot; (&#039;two&#039;) modifying &quot;mĩrongo&quot; (&#039;set-of-ten&#039;) with a nummod dependency.<br /> <br /> ==obl==<br /> <br /> The obl relation is used for non-core nominal dependents of the sentence&#039;s root. Such dependents are very common in Kikuyu, and are often describing the place of the event or other descriptors.<br /> <br /> *The phrase &quot;ndamũona kinya thaa thita ĩtigairie ndagĩka ikũmi&quot;, meaning &#039;I saw her until 10 minutes to 12&#039;, has the nominal headed by &quot;thaa&quot; (&#039;hour&#039;), which is the phrase meaning &#039;until 10 minutes to 12&#039;, with obl dependency to the verb &quot;ndamũona&quot; (I-saw), as it is not a core dependent of the verb but does depend on it.<br /> <br /> *The phrase &quot;gakinya rũũĩ-inĩ&quot;, meaning &#039;when he arrived at the river&#039;, has the locative nominal &quot;rũũĩ-inĩ&quot; (&#039;at the river&#039;) modifying the verb &quot;gakinya&quot; &#039;arrive&#039; with an obl dependency, at is a dependent on the verb but not a core dependent.<br /> <br /> ==xcomp==<br /> <br /> The xcomp relation describes the dependency between a verb&#039;s (or adjective&#039;s) subjectless clausal complement and the verb itself. Such constructions appear very often in Kikuyu, as verbs are often followed by other verbs, in analogous constructions to &quot;start to work&quot; in English. One particularly common construction involves the verb translated as &#039;return&#039;, which often takes subjectless verbal complements.<br /> <br /> *The phrase &quot;gathiĩ kuona kĩũra&quot;, meaning &#039;he went to see the frog&#039; (literally &#039;he-went see frog&#039;), has the second verb, &quot;kuona&quot; (&#039;see&#039;), with an xcomp dependency on the first verb, &quot;gathiĩ&quot; (went).<br /> <br /> *The phrase &quot;ndacoka ndanyua ũcũrũ&quot;, meaning &#039;I then drank porridge&#039; (literally I-return drink porridge), has the second verb, &quot;ndanyua&quot; (&#039;drink&#039;), with an xcomp dependency on the first verb, &quot;ndacoka&quot; (&#039;return&#039;). This is the common construction with &quot;cok&quot; (&#039;return&#039;) alluded to above, where cok plus a subjectless verbal complement is used to mean &#039;[subject] then did [verbal complement]&#039;.<br /> <br /> [[Category:sp17_UD]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Kikuyu/Universal_Dependencies&diff=4849 2017-04-25T01:35:18Z

<p>Jmundo1: </p> <hr /> <div>This page details Universal Dependencies for Kikuyu, and describes UD annotation and the training of several models on corpora.<br /> <br /> =Evaluation=<br /> <br /> <br /> Here, Corpus 1 is the corpus on which the parsers were trained, and Corpus 2 is another Kikuyu corpus. The withmorph parser was trained on the corpus with morphological information, while the nomorph parser was trained on the corpus without morphological information.<br /> <br /> The following summarizes the sizes of the corpora:<br /> {|class=&quot;wikitable&quot;<br /> |-<br /> |<br /> |Corpus 1<br /> |Corpus 2<br /> |-<br /> |Sentences<br /> |27<br /> |16<br /> |-<br /> |Forms<br /> |197<br /> |110<br /> |}<br /> <br /> The performance of the parsers on the corpora is as follows:<br /> {| class=&quot;wikitable&quot;<br /> |+Evaluation of Parsers on Corpus 1<br /> |-<br /> |<br /> |LAS<br /> |UAS<br /> |-<br /> |withmorph<br /> |60.53%<br /> |63.16%<br /> |-<br /> |nomorph<br /> |49.12%<br /> |57.02%<br /> |}<br /> <br /> For Corpus 1, the withmorph parser fared better on both LAS and UAS metrics, although the degree to which it fared better than the nomorph parser was somewhat higher for the LAS metric.<br /> <br /> {| class=&quot;wikitable&quot;<br /> |+Evaluation of Parsers on Corpus 2<br /> |-<br /> |<br /> |LAS<br /> |UAS<br /> |-<br /> |withmorph<br /> |8.86%<br /> |22.78%<br /> |-<br /> |nomorph<br /> |6.33%<br /> |30.38%<br /> |}<br /> <br /> Both parsers fared much worse on Corpus 2 than on Corpus 1, as expected, given that the parsers were trained on Corpus 1. Surprisingly, the nomorph parser actually had a better UAS score than the withmorph prsaer for Corpus 2.<br /> <br /> =Dependency Relations=<br /> <br /> In this section, five different dependency relations used in annotating the corpora are detailed.<br /> <br /> ==nmod==<br /> <br /> The nmod relation is, in general, used to mark a nominal modifying another nominal in some way. In Kikuyu, this form of modification appears often; for example, in the noun-associative-noun construction, the second nominal serves as a modifier of the first. There are also numerous constructions in which a noun modifies an adjacent noun. Examples include:<br /> <br /> *The phrase &quot;mbathi thaa inya&quot;, meaning &#039;ten o&#039;clock bus&#039; (literally &#039;bus hour four&#039;), has the noun &quot;thaa&quot; (&#039;hour&#039;) modifying the noun &quot;mbathi&quot; (&#039;bus&#039;) with an nmod relationship.<br /> <br /> *The phrase &quot;marwa ma tax&quot;, meaning &#039;tax letter&#039; (literally &#039;letter ASSOC tax&#039;), has the nominal headed by &quot;tax&quot; (&#039;tax&#039;) modifying the noun &quot;marwa&quot; (&#039;letter&#039;) with an nmod relationship.<br /> <br /> ==case==<br /> <br /> The case relation is, in general, used to mark something like a preposition which acts as a case-assigner for a nominal. In Kikuyu, we have chosen to analyze the relationship between an associative marker and the next word as one of case, as the associative plays a similar role to that of a word like &quot;of&quot; in English. Furthermore, the associative is used in possessive constructions, which usually involve a case relationship. The case relationship also appears with prepositions.<br /> <br /> *The phrase &quot;ndarĩkia na gym&quot;, meaning &#039;I finished with the gym&#039; (literally &#039;I-finished with gym&#039;), has the preposition &quot;na&quot; (&#039;with&#039;) in a case relationship with the nominal &quot;gym&quot; (&#039;gym&#039;).<br /> <br /> *The phrase &quot;Mũthenya Ũmwe wa Wambũi&quot;, meaning &#039;A Day in the Life of Wambũi&quot; (literally &#039;day one ASSOC Wambũi&#039;), has the associative &quot;wa&quot; in a case relationship with the nominal &quot;Wambũi&quot;.<br /> <br /> ==num==<br /> <br /> The num relation is used when a number is modifying a nominal. This construction appears often in Kikuyu, with the number word following the nominal.<br /> <br /> *The phrase &quot;thaa ithatũ kĩroko&quot;, meaning &#039;9 in the morning&#039; (literally &#039;hour three morning&#039;), has the number &quot;ithatũ&quot; modifying the nominal &quot;thaa&quot; (&#039;hour&#039;) with a nummod dependency. This is generally how time is expressed in Kikuyu.<br /> <br /> *The phrase &quot;mĩrongo ĩrĩ&quot;, meaning &#039;twenty&#039; (literally &#039;set-of-ten two&#039;), has the number &quot;ĩrĩ&quot; (&#039;two&#039;) modifying &quot;mĩrongo&quot; (&#039;set-of-ten&#039;) with a nummod dependency.<br /> <br /> ==obl==<br /> <br /> The obl relation is used for non-core nominal dependents of the sentence&#039;s root. Such dependents are very common in Kikuyu, and are often describing the place of the event or other descriptors.<br /> <br /> *The phrase &quot;ndamũona kinya thaa thita ĩtigairie ndagĩka ikũmi&quot;, meaning &#039;I saw her until 10 minutes to 12&#039;, has the nominal headed by &quot;thaa&quot; (&#039;hour&#039;), which is the phrase meaning &#039;until 10 minutes to 12&#039;, with obl dependency to the verb &quot;ndamũona&quot; (I-saw), as it is not a core dependent of the verb but does depend on it.<br /> <br /> *The phrase &quot;gakinya rũũĩ-inĩ&quot;, meaning &#039;when he arrived at the river&#039;, has the locative nominal &quot;rũũĩ-inĩ&quot; (&#039;at the river&#039;) modifying the verb &quot;gakinya&quot; &#039;arrive&#039; with an obl dependency, at is a dependent on the verb but not a core dependent.<br /> <br /> ==xcomp==<br /> <br /> The xcomp relation describes the dependency between a verb&#039;s (or adjective&#039;s) subjectless clausal complement and the verb itself. Such constructions appear very often in Kikuyu, as verbs are often followed by other verbs, in analogous constructions to &quot;start to work&quot; in English. One particularly common construction involves the verb translated as &#039;return&#039;, which often takes subjectless verbal complements.<br /> <br /> *The phrase &quot;gathiĩ kuona kĩũra&quot;, meaning &#039;he went to see the frog&#039; (literally &#039;he-went see frog&#039;), has the second verb, &quot;kuona&quot; (&#039;see&#039;), with an xcomp dependency on the first verb, &quot;gathiĩ&quot; (went).<br /> <br /> *The phrase &quot;ndacoka ndanyua ũcũrũ&quot;, meaning &#039;I then drank porridge&#039; (literally I-return drink porridge), has the second verb, &quot;ndanyua&quot; (&#039;drink&#039;), with an xcomp dependency on the first verb, &quot;ndacoka&quot; (&#039;return&#039;). This is the common construction with &quot;cok&quot; (&#039;return&#039;) alluded to above, where cok plus a subjectless verbal complement is used to mean &#039;[subject] then did [verbal complement]&#039;.<br /> <br /> [[Category:sp17_UD]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Kikuyu/Universal_Dependencies&diff=4848 2017-04-25T01:34:54Z

<p>Jmundo1: </p> <hr /> <div>This page details Universal Dependencies for Kikuyu, and describes UD annotation and the training of several models on corpora.<br /> <br /> =Evaluation=<br /> <br /> <br /> Here, Corpus 1 is the corpus on which the parsers were trained, and Corpus 2 is another Kikuyu corpus. The withmorph parser was trained on the corpus with morphological information, while the nomorph parser was trained on the corpus without morphological information.<br /> <br /> The following summarizes the sizes of the corpora:<br /> {|class=&quot;wikitable&quot;<br /> |-<br /> |<br /> |Corpus 1<br /> |Corpus 2<br /> |-<br /> |Sentences<br /> |27<br /> |16<br /> |-<br /> |Forms<br /> |197<br /> |110<br /> |}<br /> <br /> The performance of the parsers on the corpora is as follows:<br /> {| class=&quot;wikitable&quot;<br /> |+Evaluation of Parsers on Corpus 1<br /> |-<br /> |<br /> |LAS<br /> |UAS<br /> |-<br /> |withmorph<br /> |60.53%<br /> |63.16%<br /> |-<br /> |nomorph<br /> |49.12%<br /> |57.02%<br /> |}<br /> <br /> For Corpus 1, the withmorph parser fared better on both LAS and UAS metrics, although the degree to which it fared better than the nomorph parser was somewhat higher for the LAS metric.<br /> <br /> {| class=&quot;wikitable&quot;<br /> |+Evaluation of Parsers on Corpus 2<br /> |-<br /> |LAS<br /> |UAS<br /> |-<br /> |withmorph<br /> |8.86%<br /> |22.78%<br /> |-<br /> |nomorph<br /> |6.33%<br /> |30.38%<br /> |}<br /> <br /> Both parsers fared much worse on Corpus 2 than on Corpus 1, as expected, given that the parsers were trained on Corpus 1. Surprisingly, the nomorph parser actually had a better UAS score than the withmorph prsaer for Corpus 2.<br /> <br /> =Dependency Relations=<br /> <br /> In this section, five different dependency relations used in annotating the corpora are detailed.<br /> <br /> ==nmod==<br /> <br /> The nmod relation is, in general, used to mark a nominal modifying another nominal in some way. In Kikuyu, this form of modification appears often; for example, in the noun-associative-noun construction, the second nominal serves as a modifier of the first. There are also numerous constructions in which a noun modifies an adjacent noun. Examples include:<br /> <br /> *The phrase &quot;mbathi thaa inya&quot;, meaning &#039;ten o&#039;clock bus&#039; (literally &#039;bus hour four&#039;), has the noun &quot;thaa&quot; (&#039;hour&#039;) modifying the noun &quot;mbathi&quot; (&#039;bus&#039;) with an nmod relationship.<br /> <br /> *The phrase &quot;marwa ma tax&quot;, meaning &#039;tax letter&#039; (literally &#039;letter ASSOC tax&#039;), has the nominal headed by &quot;tax&quot; (&#039;tax&#039;) modifying the noun &quot;marwa&quot; (&#039;letter&#039;) with an nmod relationship.<br /> <br /> ==case==<br /> <br /> The case relation is, in general, used to mark something like a preposition which acts as a case-assigner for a nominal. In Kikuyu, we have chosen to analyze the relationship between an associative marker and the next word as one of case, as the associative plays a similar role to that of a word like &quot;of&quot; in English. Furthermore, the associative is used in possessive constructions, which usually involve a case relationship. The case relationship also appears with prepositions.<br /> <br /> *The phrase &quot;ndarĩkia na gym&quot;, meaning &#039;I finished with the gym&#039; (literally &#039;I-finished with gym&#039;), has the preposition &quot;na&quot; (&#039;with&#039;) in a case relationship with the nominal &quot;gym&quot; (&#039;gym&#039;).<br /> <br /> *The phrase &quot;Mũthenya Ũmwe wa Wambũi&quot;, meaning &#039;A Day in the Life of Wambũi&quot; (literally &#039;day one ASSOC Wambũi&#039;), has the associative &quot;wa&quot; in a case relationship with the nominal &quot;Wambũi&quot;.<br /> <br /> ==num==<br /> <br /> The num relation is used when a number is modifying a nominal. This construction appears often in Kikuyu, with the number word following the nominal.<br /> <br /> *The phrase &quot;thaa ithatũ kĩroko&quot;, meaning &#039;9 in the morning&#039; (literally &#039;hour three morning&#039;), has the number &quot;ithatũ&quot; modifying the nominal &quot;thaa&quot; (&#039;hour&#039;) with a nummod dependency. This is generally how time is expressed in Kikuyu.<br /> <br /> *The phrase &quot;mĩrongo ĩrĩ&quot;, meaning &#039;twenty&#039; (literally &#039;set-of-ten two&#039;), has the number &quot;ĩrĩ&quot; (&#039;two&#039;) modifying &quot;mĩrongo&quot; (&#039;set-of-ten&#039;) with a nummod dependency.<br /> <br /> ==obl==<br /> <br /> The obl relation is used for non-core nominal dependents of the sentence&#039;s root. Such dependents are very common in Kikuyu, and are often describing the place of the event or other descriptors.<br /> <br /> *The phrase &quot;ndamũona kinya thaa thita ĩtigairie ndagĩka ikũmi&quot;, meaning &#039;I saw her until 10 minutes to 12&#039;, has the nominal headed by &quot;thaa&quot; (&#039;hour&#039;), which is the phrase meaning &#039;until 10 minutes to 12&#039;, with obl dependency to the verb &quot;ndamũona&quot; (I-saw), as it is not a core dependent of the verb but does depend on it.<br /> <br /> *The phrase &quot;gakinya rũũĩ-inĩ&quot;, meaning &#039;when he arrived at the river&#039;, has the locative nominal &quot;rũũĩ-inĩ&quot; (&#039;at the river&#039;) modifying the verb &quot;gakinya&quot; &#039;arrive&#039; with an obl dependency, at is a dependent on the verb but not a core dependent.<br /> <br /> ==xcomp==<br /> <br /> The xcomp relation describes the dependency between a verb&#039;s (or adjective&#039;s) subjectless clausal complement and the verb itself. Such constructions appear very often in Kikuyu, as verbs are often followed by other verbs, in analogous constructions to &quot;start to work&quot; in English. One particularly common construction involves the verb translated as &#039;return&#039;, which often takes subjectless verbal complements.<br /> <br /> *The phrase &quot;gathiĩ kuona kĩũra&quot;, meaning &#039;he went to see the frog&#039; (literally &#039;he-went see frog&#039;), has the second verb, &quot;kuona&quot; (&#039;see&#039;), with an xcomp dependency on the first verb, &quot;gathiĩ&quot; (went).<br /> <br /> *The phrase &quot;ndacoka ndanyua ũcũrũ&quot;, meaning &#039;I then drank porridge&#039; (literally I-return drink porridge), has the second verb, &quot;ndanyua&quot; (&#039;drink&#039;), with an xcomp dependency on the first verb, &quot;ndacoka&quot; (&#039;return&#039;). This is the common construction with &quot;cok&quot; (&#039;return&#039;) alluded to above, where cok plus a subjectless verbal complement is used to mean &#039;[subject] then did [verbal complement]&#039;.<br /> <br /> [[Category:sp17_UD]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Kikuyu/Universal_Dependencies&diff=4847 2017-04-25T01:34:16Z

<p>Jmundo1: /* Evaluation */</p> <hr /> <div>This page details Universal Dependencies for Kikuyu, and describes UD annotation and the training of several models on corpora.<br /> <br /> =Evaluation=<br /> <br /> <br /> Here, Corpus 1 is the corpus on which the parsers were trained, and Corpus 2 is another Kikuyu corpus. The withmorph parser was trained on the corpus with morphological information, while the nomorph parser was trained on the corpus without morphological information.<br /> <br /> The following summarizes the sizes of the corpora:<br /> {|class=&quot;wikitable&quot;<br /> |-<br /> |<br /> |Corpus 1<br /> |Corpus 2<br /> |-<br /> |Sentences<br /> |27<br /> |16<br /> |-<br /> |Forms<br /> |197<br /> |110<br /> |}<br /> <br /> The performance of the parsers on the corpora is as follows:<br /> {| class=&quot;wikitable&quot;<br /> |+Evaluation of Parsers on Corpus 1<br /> |-<br /> |<br /> |LAS<br /> |UAS<br /> |-<br /> |withmorph<br /> |60.53%<br /> |63.16%<br /> |-<br /> |nomorph<br /> |49.12%<br /> |57.02%<br /> |}<br /> <br /> For Corpus 1, the withmorph parser fared better on both LAS and UAS metrics, although the degree to which it fared better than the nomorph parser was somewhat higher for the LAS metric.<br /> <br /> {| class=&quot;wikitable&quot;<br /> |+Evaluation of Parsers on Corpus 2<br /> |-<br /> |LAS<br /> |UAS<br /> |-<br /> |withmorph<br /> |8.86%<br /> |22.78%<br /> |-<br /> |nomorph<br /> |6.33%<br /> |30.38%<br /> |}<br /> <br /> Both parsers fared much worse on Corpus 2 than on Corpus 1, as expected, given that the parsers were trained on Corpus 1. Surprisingly, the nomorph parser actually had a better UAS score than the withmorph prsaer for Corpus 2.<br /> <br /> =Dependency Relations=<br /> <br /> In this section, five different dependency relations used in annotating the corpora are detailed.<br /> <br /> ==nmod==<br /> <br /> The nmod relation is, in general, used to mark a nominal modifying another nominal in some way. In Kikuyu, this form of modification appears often; for example, in the noun-associative-noun construction, the second nominal serves as a modifier of the first. There are also numerous constructions in which a noun modifies an adjacent noun. Examples include:<br /> <br /> *The phrase &quot;mbathi thaa inya&quot;, meaning &#039;ten o&#039;clock bus&#039; (literally &#039;bus hour four<br /> ), has the noun &quot;thaa&quot; (&#039;hour&#039;) modifying the noun &quot;mbathi&quot; (&#039;bus&#039;) with an nmod relationship.<br /> <br /> *The phrase &quot;marwa ma tax&quot;, meaning &#039;tax letter&#039; (literally &#039;letter ASSOC tax&#039;), has the nominal headed by &quot;tax&quot; (&#039;tax&#039;) modifying the noun &quot;marwa&quot; (&#039;letter&#039;) with an nmod relationship.<br /> <br /> ==case==<br /> <br /> The case relation is, in general, used to mark something like a preposition which acts as a case-assigner for a nominal. In Kikuyu, we have chosen to analyze the relationship between an associative marker and the next word as one of case, as the associative plays a similar role to that of a word like &quot;of&quot; in English. Furthermore, the associative is used in possessive constructions, which usually involve a case relationship. The case relationship also appears with prepositions.<br /> <br /> *The phrase &quot;ndarĩkia na gym&quot;, meaning &#039;I finished with the gym&#039; (literally &#039;I-finished with gym&#039;), has the preposition &quot;na&quot; (&#039;with&#039;) in a case relationship with the nominal &quot;gym&quot; (&#039;gym&#039;).<br /> <br /> *The phrase &quot;Mũthenya Ũmwe wa Wambũi&quot;, meaning &#039;A Day in the Life of Wambũi&quot; (literally &#039;day one ASSOC Wambũi&#039;), has the associative &quot;wa&quot; in a case relationship with the nominal &quot;Wambũi&quot;.<br /> <br /> ==num==<br /> <br /> The num relation is used when a number is modifying a nominal. This construction appears often in Kikuyu, with the number word following the nominal.<br /> <br /> *The phrase &quot;thaa ithatũ kĩroko&quot;, meaning &#039;9 in the morning&#039; (literally &#039;hour three morning&#039;), has the number &quot;ithatũ&quot; modifying the nominal &quot;thaa&quot; (&#039;hour&#039;) with a nummod dependency. This is generally how time is expressed in Kikuyu.<br /> <br /> *The phrase &quot;mĩrongo ĩrĩ&quot;, meaning &#039;twenty&#039; (literally &#039;set-of-ten two&#039;), has the number &quot;ĩrĩ&quot; (&#039;two&#039;) modifying &quot;mĩrongo&quot; (&#039;set-of-ten&#039;) with a nummod dependency.<br /> <br /> ==obl==<br /> <br /> The obl relation is used for non-core nominal dependents of the sentence&#039;s root. Such dependents are very common in Kikuyu, and are often describing the place of the event or other descriptors.<br /> <br /> *The phrase &quot;ndamũona kinya thaa thita ĩtigairie ndagĩka ikũmi&quot;, meaning &#039;I saw her until 10 minutes to 12&#039;, has the nominal headed by &quot;thaa&quot; (&#039;hour&#039;), which is the phrase meaning &#039;until 10 minutes to 12&#039;, with obl dependency to the verb &quot;ndamũona&quot; (I-saw), as it is not a core dependent of the verb but does depend on it.<br /> <br /> *The phrase &quot;gakinya rũũĩ-inĩ&quot;, meaning &#039;when he arrived at the river&#039;, has the locative nominal &quot;rũũĩ-inĩ&quot; (&#039;at the river&#039;) modifying the verb &quot;gakinya&quot; &#039;arrive&#039; with an obl dependency, at is a dependent on the verb but not a core dependent.<br /> <br /> ==xcomp==<br /> <br /> The xcomp relation describes the dependency between a verb&#039;s (or adjective&#039;s) subjectless clausal complement and the verb itself. Such constructions appear very often in Kikuyu, as verbs are often followed by other verbs, in analogous constructions to &quot;start to work&quot; in English. One particularly common construction involves the verb translated as &#039;return&#039;, which often takes subjectless verbal complements.<br /> <br /> *The phrase &quot;gathiĩ kuona kĩũra&quot;, meaning &#039;he went to see the frog&#039; (literally &#039;he-went see frog&#039;), has the second verb, &quot;kuona&quot; (&#039;see&#039;), with an xcomp dependency on the first verb, &quot;gathiĩ&quot; (went).<br /> <br /> *The phrase &quot;ndacoka ndanyua ũcũrũ&quot;, meaning &#039;I then drank porridge&#039; (literally I-return drink porridge), has the second verb, &quot;ndanyua&quot; (&#039;drink&#039;), with an xcomp dependency on the first verb, &quot;ndacoka&quot; (&#039;return&#039;). This is the common construction with &quot;cok&quot; (&#039;return&#039;) alluded to above, where cok plus a subjectless verbal complement is used to mean &#039;[subject] then did [verbal complement]&#039;.<br /> <br /> [[Category:sp17_UD]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Kikuyu/Universal_Dependencies&diff=4846 2017-04-25T01:32:05Z

<p>Jmundo1: </p> <hr /> <div>This page details Universal Dependencies for Kikuyu, and describes UD annotation and the training of several models on corpora.<br /> <br /> =Evaluation=<br /> <br /> <br /> Here, Corpus 1 is the corpus on which the parsers were trained, and Corpus 2 is another Kikuyu corpus. The withmorph parser was trained on the corpus with morphological information, while the nomorph parser was trained on the corpus without morphological information.<br /> {| class=&quot;wikitable&quot;<br /> |+Evaluation of Parsers on Corpus 1 (27 sentences, 197 forms)<br /> |-<br /> |<br /> |LAS<br /> |UAS<br /> |-<br /> |withmorph<br /> |60.53%<br /> |63.16%<br /> |nomorph<br /> |49.12%<br /> |57.02%<br /> |}<br /> <br /> For Corpus 1, the withmorph parser fared better on both LAS and UAS metrics, although the degree to which it fared better than the nomorph parser was somewhat higher for the LAS metric.<br /> <br /> {| class=&quot;wikitable&quot;<br /> |+Evaluation of Parsers on Corpus 2 (16 sentences, 110 forms)<br /> |-<br /> |LAS<br /> |UAS<br /> |-<br /> |withmorph<br /> |8.86%<br /> |22.78%<br /> |nomorph<br /> |6.33%<br /> |30.38%<br /> |}<br /> <br /> Both parsers fared much worse on Corpus 2 than on Corpus 1, as expected, given that the parsers were trained on Corpus 1. Surprisingly, the nomorph parser actually had a better UAS score than the withmorph prsaer for Corpus 2.<br /> <br /> <br /> =Dependency Relations=<br /> <br /> In this section, five different dependency relations used in annotating the corpora are detailed.<br /> <br /> ==nmod==<br /> <br /> The nmod relation is, in general, used to mark a nominal modifying another nominal in some way. In Kikuyu, this form of modification appears often; for example, in the noun-associative-noun construction, the second nominal serves as a modifier of the first. There are also numerous constructions in which a noun modifies an adjacent noun. Examples include:<br /> <br /> *The phrase &quot;mbathi thaa inya&quot;, meaning &#039;ten o&#039;clock bus&#039; (literally &#039;bus hour four<br /> ), has the noun &quot;thaa&quot; (&#039;hour&#039;) modifying the noun &quot;mbathi&quot; (&#039;bus&#039;) with an nmod relationship.<br /> <br /> *The phrase &quot;marwa ma tax&quot;, meaning &#039;tax letter&#039; (literally &#039;letter ASSOC tax&#039;), has the nominal headed by &quot;tax&quot; (&#039;tax&#039;) modifying the noun &quot;marwa&quot; (&#039;letter&#039;) with an nmod relationship.<br /> <br /> ==case==<br /> <br /> The case relation is, in general, used to mark something like a preposition which acts as a case-assigner for a nominal. In Kikuyu, we have chosen to analyze the relationship between an associative marker and the next word as one of case, as the associative plays a similar role to that of a word like &quot;of&quot; in English. Furthermore, the associative is used in possessive constructions, which usually involve a case relationship. The case relationship also appears with prepositions.<br /> <br /> *The phrase &quot;ndarĩkia na gym&quot;, meaning &#039;I finished with the gym&#039; (literally &#039;I-finished with gym&#039;), has the preposition &quot;na&quot; (&#039;with&#039;) in a case relationship with the nominal &quot;gym&quot; (&#039;gym&#039;).<br /> <br /> *The phrase &quot;Mũthenya Ũmwe wa Wambũi&quot;, meaning &#039;A Day in the Life of Wambũi&quot; (literally &#039;day one ASSOC Wambũi&#039;), has the associative &quot;wa&quot; in a case relationship with the nominal &quot;Wambũi&quot;.<br /> <br /> ==num==<br /> <br /> The num relation is used when a number is modifying a nominal. This construction appears often in Kikuyu, with the number word following the nominal.<br /> <br /> *The phrase &quot;thaa ithatũ kĩroko&quot;, meaning &#039;9 in the morning&#039; (literally &#039;hour three morning&#039;), has the number &quot;ithatũ&quot; modifying the nominal &quot;thaa&quot; (&#039;hour&#039;) with a nummod dependency. This is generally how time is expressed in Kikuyu.<br /> <br /> *The phrase &quot;mĩrongo ĩrĩ&quot;, meaning &#039;twenty&#039; (literally &#039;set-of-ten two&#039;), has the number &quot;ĩrĩ&quot; (&#039;two&#039;) modifying &quot;mĩrongo&quot; (&#039;set-of-ten&#039;) with a nummod dependency.<br /> <br /> ==obl==<br /> <br /> The obl relation is used for non-core nominal dependents of the sentence&#039;s root. Such dependents are very common in Kikuyu, and are often describing the place of the event or other descriptors.<br /> <br /> *The phrase &quot;ndamũona kinya thaa thita ĩtigairie ndagĩka ikũmi&quot;, meaning &#039;I saw her until 10 minutes to 12&#039;, has the nominal headed by &quot;thaa&quot; (&#039;hour&#039;), which is the phrase meaning &#039;until 10 minutes to 12&#039;, with obl dependency to the verb &quot;ndamũona&quot; (I-saw), as it is not a core dependent of the verb but does depend on it.<br /> <br /> *The phrase &quot;gakinya rũũĩ-inĩ&quot;, meaning &#039;when he arrived at the river&#039;, has the locative nominal &quot;rũũĩ-inĩ&quot; (&#039;at the river&#039;) modifying the verb &quot;gakinya&quot; &#039;arrive&#039; with an obl dependency, at is a dependent on the verb but not a core dependent.<br /> <br /> ==xcomp==<br /> <br /> The xcomp relation describes the dependency between a verb&#039;s (or adjective&#039;s) subjectless clausal complement and the verb itself. Such constructions appear very often in Kikuyu, as verbs are often followed by other verbs, in analogous constructions to &quot;start to work&quot; in English. One particularly common construction involves the verb translated as &#039;return&#039;, which often takes subjectless verbal complements.<br /> <br /> *The phrase &quot;gathiĩ kuona kĩũra&quot;, meaning &#039;he went to see the frog&#039; (literally &#039;he-went see frog&#039;), has the second verb, &quot;kuona&quot; (&#039;see&#039;), with an xcomp dependency on the first verb, &quot;gathiĩ&quot; (went).<br /> <br /> *The phrase &quot;ndacoka ndanyua ũcũrũ&quot;, meaning &#039;I then drank porridge&#039; (literally I-return drink porridge), has the second verb, &quot;ndanyua&quot; (&#039;drink&#039;), with an xcomp dependency on the first verb, &quot;ndacoka&quot; (&#039;return&#039;). This is the common construction with &quot;cok&quot; (&#039;return&#039;) alluded to above, where cok plus a subjectless verbal complement is used to mean &#039;[subject] then did [verbal complement]&#039;.<br /> <br /> [[Category:sp17_UD]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Kikuyu/Universal_Dependencies&diff=4845 2017-04-25T01:31:44Z

<p>Jmundo1: </p> <hr /> <div>This page details Universal Dependencies for Kikuyu, and describes UD annotation and the training of several models on corpora.<br /> <br /> =Evaluation=<br /> <br /> <br /> ==Evaluation==<br /> Here, Corpus 1 is the corpus on which the parsers were trained, and Corpus 2 is another Kikuyu corpus. The withmorph parser was trained on the corpus with morphological information, while the nomorph parser was trained on the corpus without morphological information.<br /> {| class=&quot;wikitable&quot;<br /> |+Evaluation of Parsers on Corpus 1 (27 sentences, 197 forms)<br /> |-<br /> |<br /> |LAS<br /> |UAS<br /> |-<br /> |withmorph<br /> |60.53%<br /> |63.16%<br /> |nomorph<br /> |49.12%<br /> |57.02%<br /> |}<br /> <br /> For Corpus 1, the withmorph parser fared better on both LAS and UAS metrics, although the degree to which it fared better than the nomorph parser was somewhat higher for the LAS metric.<br /> <br /> {| class=&quot;wikitable&quot;<br /> |+Evaluation of Parsers on Corpus 2 (16 sentences, 110 forms)<br /> |-<br /> |LAS<br /> |UAS<br /> |-<br /> |withmorph<br /> |8.86%<br /> |22.78%<br /> |nomorph<br /> |6.33%<br /> |30.38%<br /> |}<br /> <br /> Both parsers fared much worse on Corpus 2 than on Corpus 1, as expected, given that the parsers were trained on Corpus 1. Surprisingly, the nomorph parser actually had a better UAS score than the withmorph prsaer for Corpus 2.<br /> <br /> <br /> =Dependency Relations=<br /> <br /> In this section, five different dependency relations used in annotating the corpora are detailed.<br /> <br /> ==nmod==<br /> <br /> The nmod relation is, in general, used to mark a nominal modifying another nominal in some way. In Kikuyu, this form of modification appears often; for example, in the noun-associative-noun construction, the second nominal serves as a modifier of the first. There are also numerous constructions in which a noun modifies an adjacent noun. Examples include:<br /> <br /> *The phrase &quot;mbathi thaa inya&quot;, meaning &#039;ten o&#039;clock bus&#039; (literally &#039;bus hour four<br /> ), has the noun &quot;thaa&quot; (&#039;hour&#039;) modifying the noun &quot;mbathi&quot; (&#039;bus&#039;) with an nmod relationship.<br /> <br /> *The phrase &quot;marwa ma tax&quot;, meaning &#039;tax letter&#039; (literally &#039;letter ASSOC tax&#039;), has the nominal headed by &quot;tax&quot; (&#039;tax&#039;) modifying the noun &quot;marwa&quot; (&#039;letter&#039;) with an nmod relationship.<br /> <br /> ==case==<br /> <br /> The case relation is, in general, used to mark something like a preposition which acts as a case-assigner for a nominal. In Kikuyu, we have chosen to analyze the relationship between an associative marker and the next word as one of case, as the associative plays a similar role to that of a word like &quot;of&quot; in English. Furthermore, the associative is used in possessive constructions, which usually involve a case relationship. The case relationship also appears with prepositions.<br /> <br /> *The phrase &quot;ndarĩkia na gym&quot;, meaning &#039;I finished with the gym&#039; (literally &#039;I-finished with gym&#039;), has the preposition &quot;na&quot; (&#039;with&#039;) in a case relationship with the nominal &quot;gym&quot; (&#039;gym&#039;).<br /> <br /> *The phrase &quot;Mũthenya Ũmwe wa Wambũi&quot;, meaning &#039;A Day in the Life of Wambũi&quot; (literally &#039;day one ASSOC Wambũi&#039;), has the associative &quot;wa&quot; in a case relationship with the nominal &quot;Wambũi&quot;.<br /> <br /> ==num==<br /> <br /> The num relation is used when a number is modifying a nominal. This construction appears often in Kikuyu, with the number word following the nominal.<br /> <br /> *The phrase &quot;thaa ithatũ kĩroko&quot;, meaning &#039;9 in the morning&#039; (literally &#039;hour three morning&#039;), has the number &quot;ithatũ&quot; modifying the nominal &quot;thaa&quot; (&#039;hour&#039;) with a nummod dependency. This is generally how time is expressed in Kikuyu.<br /> <br /> *The phrase &quot;mĩrongo ĩrĩ&quot;, meaning &#039;twenty&#039; (literally &#039;set-of-ten two&#039;), has the number &quot;ĩrĩ&quot; (&#039;two&#039;) modifying &quot;mĩrongo&quot; (&#039;set-of-ten&#039;) with a nummod dependency.<br /> <br /> ==obl==<br /> <br /> The obl relation is used for non-core nominal dependents of the sentence&#039;s root. Such dependents are very common in Kikuyu, and are often describing the place of the event or other descriptors.<br /> <br /> *The phrase &quot;ndamũona kinya thaa thita ĩtigairie ndagĩka ikũmi&quot;, meaning &#039;I saw her until 10 minutes to 12&#039;, has the nominal headed by &quot;thaa&quot; (&#039;hour&#039;), which is the phrase meaning &#039;until 10 minutes to 12&#039;, with obl dependency to the verb &quot;ndamũona&quot; (I-saw), as it is not a core dependent of the verb but does depend on it.<br /> <br /> *The phrase &quot;gakinya rũũĩ-inĩ&quot;, meaning &#039;when he arrived at the river&#039;, has the locative nominal &quot;rũũĩ-inĩ&quot; (&#039;at the river&#039;) modifying the verb &quot;gakinya&quot; &#039;arrive&#039; with an obl dependency, at is a dependent on the verb but not a core dependent.<br /> <br /> ==xcomp==<br /> <br /> The xcomp relation describes the dependency between a verb&#039;s (or adjective&#039;s) subjectless clausal complement and the verb itself. Such constructions appear very often in Kikuyu, as verbs are often followed by other verbs, in analogous constructions to &quot;start to work&quot; in English. One particularly common construction involves the verb translated as &#039;return&#039;, which often takes subjectless verbal complements.<br /> <br /> *The phrase &quot;gathiĩ kuona kĩũra&quot;, meaning &#039;he went to see the frog&#039; (literally &#039;he-went see frog&#039;), has the second verb, &quot;kuona&quot; (&#039;see&#039;), with an xcomp dependency on the first verb, &quot;gathiĩ&quot; (went).<br /> <br /> *The phrase &quot;ndacoka ndanyua ũcũrũ&quot;, meaning &#039;I then drank porridge&#039; (literally I-return drink porridge), has the second verb, &quot;ndanyua&quot; (&#039;drink&#039;), with an xcomp dependency on the first verb, &quot;ndacoka&quot; (&#039;return&#039;). This is the common construction with &quot;cok&quot; (&#039;return&#039;) alluded to above, where cok plus a subjectless verbal complement is used to mean &#039;[subject] then did [verbal complement]&#039;.<br /> <br /> [[Category:sp17_UD]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Kikuyu/Universal_Dependencies&diff=4827 2017-04-25T00:11:08Z

<p>Jmundo1: </p> <hr /> <div><br /> =Dependency Relations=<br /> <br /> In this section, five different dependency relations used in annotating the corpora are detailed.<br /> <br /> ==nmod==<br /> <br /> The nmod relation is, in general, used to mark a nominal modifying another nominal in some way. In Kikuyu, this form of modification appears often; for example, in the noun-associative-noun construction, the second nominal serves as a modifier of the first. There are also numerous constructions in which a noun modifies an adjacent noun. Examples include:<br /> <br /> *The phrase &quot;mbathi thaa inya&quot;, meaning &#039;ten o&#039;clock bus&#039; (literally &#039;bus hour four<br /> ), has the noun &quot;thaa&quot; (&#039;hour&#039;) modifying the noun &quot;mbathi&quot; (&#039;bus&#039;) with an nmod relationship.<br /> <br /> *The phrase &quot;marwa ma tax&quot;, meaning &#039;tax letter&#039; (literally &#039;letter ASSOC tax&#039;), has the nominal headed by &quot;tax&quot; (&#039;tax&#039;) modifying the noun &quot;marwa&quot; (&#039;letter&#039;) with an nmod relationship.<br /> <br /> ==case==<br /> <br /> The case relation is, in general, used to mark something like a preposition which acts as a case-assigner for a nominal. In Kikuyu, we have chosen to analyze the relationship between an associative marker and the next word as one of case, as the associative plays a similar role to that of a word like &quot;of&quot; in English. Furthermore, the associative is used in possessive constructions, which usually involve a case relationship. The case relationship also appears with prepositions.<br /> <br /> *The phrase &quot;ndarĩkia na gym&quot;, meaning &#039;I finished with the gym&#039; (literally &#039;I-finished with gym&#039;), has the preposition &quot;na&quot; (&#039;with&#039;) in a case relationship with the nominal &quot;gym&quot; (&#039;gym&#039;).<br /> <br /> *The phrase &quot;Mũthenya Ũmwe wa Wambũi&quot;, meaning &#039;A Day in the Life of Wambũi&quot; (literally &#039;day one ASSOC Wambũi&#039;), has the associative &quot;wa&quot; in a case relationship with the nominal &quot;Wambũi&quot;.<br /> <br /> ==num==<br /> <br /> The num relation is used when a number is modifying a nominal. This construction appears often in Kikuyu, with the number word following the nominal.<br /> <br /> *The phrase &quot;thaa ithatũ kĩroko&quot;, meaning &#039;9 in the morning&#039; (literally &#039;hour three morning&#039;), has the number &quot;ithatũ&quot; modifying the nominal &quot;thaa&quot; (&#039;hour&#039;) with a nummod dependency. This is generally how time is expressed in Kikuyu.<br /> <br /> *The phrase &quot;mĩrongo ĩrĩ&quot;, meaning &#039;twenty&#039; (literally &#039;set-of-ten two&#039;), has the number &quot;ĩrĩ&quot; (&#039;two&#039;) modifying &quot;mĩrongo&quot; (&#039;set-of-ten&#039;) with a nummod dependency.<br /> <br /> ==obl==<br /> <br /> The obl relation is used for non-core nominal dependents of the sentence&#039;s root. Such dependents are very common in Kikuyu, and are often describing the place of the event or other descriptors.<br /> <br /> *The phrase &quot;ndamũona kinya thaa thita ĩtigairie ndagĩka ikũmi&quot;, meaning &#039;I saw her until 10 minutes to 12&#039;, has the nominal headed by &quot;thaa&quot; (&#039;hour&#039;), which is the phrase meaning &#039;until 10 minutes to 12&#039;, with obl dependency to the verb &quot;ndamũona&quot; (I-saw), as it is not a core dependent of the verb but does depend on it.<br /> <br /> *The phrase &quot;gakinya rũũĩ-inĩ&quot;, meaning &#039;when he arrived at the river&#039;, has the locative nominal &quot;rũũĩ-inĩ&quot; (&#039;at the river&#039;) modifying the verb &quot;gakinya&quot; &#039;arrive&#039; with an obl dependency, at is a dependent on the verb but not a core dependent.<br /> <br /> ==xcomp==<br /> <br /> The xcomp relation describes the dependency between a verb&#039;s (or adjective&#039;s) subjectless clausal complement and the verb itself. Such constructions appear very often in Kikuyu, as verbs are often followed by other verbs, in analogous constructions to &quot;start to work&quot; in English. One particularly common construction involves the verb translated as &#039;return&#039;, which often takes subjectless verbal complements.<br /> <br /> *The phrase &quot;gathiĩ kuona kĩũra&quot;, meaning &#039;he went to see the frog&#039; (literally &#039;he-went see frog&#039;), has the second verb, &quot;kuona&quot; (&#039;see&#039;), with an xcomp dependency on the first verb, &quot;gathiĩ&quot; (went).<br /> <br /> *The phrase &quot;ndacoka ndanyua ũcũrũ&quot;, meaning &#039;I then drank porridge&#039; (literally I-return drink porridge), has the second verb, &quot;ndanyua&quot; (&#039;drink&#039;), with an xcomp dependency on the first verb, &quot;ndacoka&quot; (&#039;return&#039;). This is the common construction with &quot;cok&quot; (&#039;return&#039;) alluded to above, where cok plus a subjectless verbal complement is used to mean &#039;[subject] then did [verbal complement]&#039;.<br /> <br /> [[Category:sp17_UD]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Kikuyu/Universal_Dependencies&diff=4826 2017-04-25T00:10:11Z

<p>Jmundo1: </p> <hr /> <div><br /> =Dependency Relations=<br /> <br /> In this section, five different dependency relations used in annotating the corpora are detailed.<br /> <br /> ==nmod==<br /> <br /> The nmod relation is, in general, used to mark a nominal modifying another nominal in some way. In Kikuyu, this form of modification appears often; for example, in the noun-associative-noun construction, the second nominal serves as a modifier of the first. There are also numerous constructions in which a noun modifies an adjacent noun. Examples include:<br /> <br /> *The phrase &quot;mbathi thaa inya&quot;, meaning &#039;ten o&#039;clock bus&#039; (literally &#039;bus hour four<br /> ), has the noun &quot;thaa&quot; (&#039;hour&#039;) modifying the noun &quot;mbathi&quot; (&#039;bus&#039;) with an nmod relationship.<br /> <br /> *The phrase &quot;marwa ma tax&quot;, meaning &#039;tax letter&#039; (literally &#039;letter ASSOC tax&#039;), has the nominal headed by &quot;tax&quot; (&#039;tax&#039;) modifying the noun &quot;marwa&quot; (&#039;letter&#039;) with an nmod relationship.<br /> <br /> ==case==<br /> <br /> The case relation is, in general, used to mark something like a preposition which acts as a case-assigner for a nominal. In Kikuyu, we have chosen to analyze the relationship between an associative marker and the next word as one of case, as the associative plays a similar role to that of a word like &quot;of&quot; in English. Furthermore, the associative is used in possessive constructions, which usually involve a case relationship. The case relationship also appears with prepositions.<br /> <br /> *The phrase &quot;ndarĩkia na gym&quot;, meaning &#039;I finished with the gym&#039; (literally &#039;I-finished with gym&#039;), has the preposition &quot;na&quot; (&#039;with&#039;) in a case relationship with the nominal &quot;gym&quot; (&#039;gym&#039;).<br /> <br /> *The phrase &quot;Mũthenya Ũmwe wa Wambũi&quot;, meaning &#039;A Day in the Life of Wambũi&quot; (literally &#039;day one ASSOC Wambũi&#039;), has the associative &quot;wa&quot; in a case relationship with the nominal &quot;Wambũi&quot;.<br /> <br /> ==num==<br /> <br /> The num relation is used when a number is modifying a nominal. This construction appears often in Kikuyu, with the number word following the nominal.<br /> <br /> *The phrase &quot;thaa ithatũ kĩroko&quot;, meaning &#039;9 in the morning&#039; (literally &#039;hour three morning&#039;), has the number &quot;ithatũ&quot; modifying the nominal &quot;thaa&quot; (&#039;hour&#039;) with a nummod dependency. This is generally how time is expressed in Kikuyu.<br /> <br /> *The phrase &quot;mĩrongo ĩrĩ&quot;, meaning &#039;twenty&#039; (literally &#039;set-of-ten two&#039;), has the number &quot;ĩrĩ&quot; (&#039;two&#039;) modifying &quot;mĩrongo&quot; (&#039;set-of-ten&#039;) with a nummod dependency.<br /> <br /> ==obl==<br /> <br /> The obl relation is used for non-core nominal dependents of the sentence&#039;s root. Such dependents are very common in Kikuyu, and are often describing the place of the event or other descriptors.<br /> <br /> *The phrase &quot;ndamũona kinya thaa thita ĩtigairie ndagĩka ikũmi&quot;, meaning &#039;I saw her until 10 minutes to 12&#039;, has the nominal headed by &quot;thaa&quot; (&#039;hour&#039;), which is the phrase meaning &#039;until 10 minutes to 12&#039;, with obl dependency to the verb &quot;ndamũona&quot; (I-saw), as it is not a core dependent of the verb but does depend on it.<br /> <br /> *The phrase &quot;gakinya rũũĩ-inĩ&quot;, meaning &#039;when he arrived at the river&#039;, has the locative nominal &quot;rũũĩ-inĩ&quot; (&#039;at the river&#039;) modifying the verb &quot;gakinya&quot; &#039;arrive&#039; with an obl dependency, at is a dependent on the verb but not a core dependent.<br /> <br /> ==xcomp==<br /> <br /> The xcomp relation describes the dependency between a verb&#039;s (or adjective&#039;s) subjectless clausal complement and the verb itself. Such constructions appear very often in Kikuyu, as verbs are often followed by other verbs, in analogous constructions to &quot;start to work&quot; in English. One particularly common construction involves the verb translated as &#039;return&#039;, which often takes subjectless verbal complements.<br /> <br /> *The phrase &quot;gathiĩ kuona kĩũra&quot;, meaning <br /> &#039;he went to see the frog&#039; (literally &#039;he-went see frog&#039;), has the second verb, &quot;kuona&quot; (&#039;see&#039;), with an xcomp dependency on the first verb, &quot;gathiĩ&quot; (went).<br /> <br /> *The phrase &quot;ndacoka ndanyua ũcũrũ&quot;, meaning &#039;I then drank porridge&#039; (literally I-return drink porridge), has the second verb, &quot;ndanyua&quot; (&#039;drink&#039;), with an xcomp dependency on the first verb, &quot;ndacoka&quot; (&#039;return&#039;). This is the common construction with &quot;cok&quot; (&#039;return&#039;) alluded to above, where cok plus a subjectless verbal complement is used to mean &#039;[subject] then did [verbal complement]&#039;.<br /> <br /> [[Category:sp17_UD]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Kikuyu&diff=4821 2017-04-24T23:41:49Z

<p>Jmundo1: </p> <hr /> <div>==Resources==<br /> <br /> ===Computational Resources===<br /> <br /> ====Existing Resources====<br /> <br /> *[http://extensions.services.openoffice.org/en/project/Gikuyu A beta version of a spell-checker] is downloadable online, as well as [https://www.researchgate.net/publication/228831175_Developing_an_Open_Source_Spellchecker_for_Gikuyu an article] on the development of this spell-checker. It does not appear to have been updated since 2010, and the site warns it may not be compatible with current versions of OpenOffice.<br /> <br /> ====Developed Resources====<br /> <br /> *Information on keyboard layouts for Kikuyu and to IPA in Kikuyu can be found [https://wikis.swarthmore.edu/ling073/Kikuyu/Keyboard here].<br /> <br /> *Information on Kikuyu morphology can be fond [https://wikis.swarthmore.edu/ling073/Kikuyu/Grammar here].<br /> <br /> *A [https://wikis.swarthmore.edu/ling073/Kikuyu/Transducer work-in-progress transducer] can be downloaded from [https://github.swarthmore.edu/jmundo1/ling073-kik Swarthmore&#039;s Github].<br /> <br /> *Information about the disambiguator can be found [https://wikis.swarthmore.edu/ling073/Kikuyu/Disambiguation here].<br /> <br /> *Universal dependencies tools can be found at [[Kikuyu/Universal Dependencies]].<br /> <br /> ===Dictionaries===<br /> <br /> *[https://archive.org/details/cu31924026919815 A vocabulary of Kikuyu], translating between Kikuyu and English, has been compiled in a book from 1904. It also contains information on several other languages, Swahili and Kamba. The orthography is not today&#039;s standard.<br /> <br /> *An [https://glosbe.com/en/ki/ English to Kikuyu] dictionary exists online, but while it has a few words it appears to be very few; it cannot, for example, produce reliable translations for &quot;cat&quot; or &quot;hello&quot;. It claims it has &quot;5729350 sentences translated&quot; but it is not clear what that means. It appears that anyone can edit the dictionary to add new words, and it claims that the dictionary does not store words but rather &quot;the idea of what the word means&quot;; it is also unclear what that means exactly. Furthermore, it does not appear to use the standard Kikuyu orthography for the few words it does have.<br /> <br /> ===Grammars===<br /> <br /> *[http://www.ruf.rice.edu/~reng/kik/sketch.pdf A fairly detailed grammar] of Kikuyu is available online, compiled by a 2014-2015 Rice University Field Methods course. The appendix includes quite a few examples of glossed Kikuyu data. [http://www.ruf.rice.edu/~reng/kik Sound files of these data] are also available, licensed under Creative Commons, Attribution-NonCommercial 3.0 Unported License.<br /> <br /> *[http://www.omniglot.com/writing/kikuyu.htm Basic information about Kikuyu&#039;s alphabet and orthography] can be found online.<br /> <br /> ===Scientific Works===<br /> <br /> *Kamu Chege et. al. at the University of Nairobi and the University of Antwerp have written [https://www.researchgate.net/publication/228831175_Developing_an_Open_Source_Spellchecker_for_Gikuyu an article] describing the development of an open-source spellchecker for Kikuyu. <br /> <br /> *John J. McCarthy et. al. at the University of Massachusetts - Amherst have written [https://works.bepress.com/john_j_mccarthy/109/ a phonology paper] about tone in Kikuyu.<br /> <br /> *Seth Cable has written [http://ling.auf.net/lingbuzz/001374 multiple] [http://ling.auf.net/lingbuzz/001536 papers] about tense and aspect in Kikuyu.<br /> ===Corpora===<br /> <br /> *We have compiled material from several of the corpora below. The repository cannot be made public due to licensing issues but can be found on [https://github.swarthmore.edu/jmundo1/ling073-kik-corpus Swarthmore&#039;s github] for those with access.<br /> <br /> *[https://ki.wikipedia.org/wiki/Main_Page Kikuyu Wikipedia] exists, but it appears to be somewhat limited, with a bit over 1000 articles. Text is available under the Creative Commons Attribution-ShareAlike License.<br /> <br /> *[http://gigikuyu.blogspot.com/ Several] [http://gikuyukiamugikuyu.blogspot.com/ online] [http://gikuyu.blogspot.com/ blogs] in Kikuyu exist and are available. It is hard to tell what they are about. [http://gikuyukiamugikuyu.blogspot.com/ This blog] was used for the development of the open-source Kikuyu spellchecker mentioned above. There is no apparent licensing of any of these blogs.<br /> <br /> *[https://www.bible.com/bible/1202/jhn.1.gikdc The bible] has been translated into Kikuyu and is also available [https://play.google.com/store/apps/details?id=kikuyu.holy.bible.AOVGUEHTERZFTWPYRJ&amp;hl=en as an app].<br /> <br /> *[http://www.mutiri.com/ This website] appears to be entirely in Kikuyu, and it is not clear what it is, other than that it has [http://www.mutiri.com/njaranda/82009/writers.shtml a page] listing what appear to be biographies of people. The website appears to be Copyright 2007-2010 Muutiiri.com, but it&#039;s hard to be certain without being able to read Kikuyu.<br /> <br /> *[https://issuu.com/m.ngethe/docs/dan_mag_finale_april_insides A Kikuyu magazine of some sort] is available online.<br /> <br /> [[Category:sp17_ResourceDocumentation]]<br /> [[Category:Kikuyu]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Kikuyu/Universal_Dependencies&diff=4820 2017-04-24T23:41:42Z

<p>Jmundo1: Created page with &quot;a&quot;</p> <hr /> <div>a</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Lingala_and_Kikuyu&diff=4593 2017-04-18T02:48:47Z

<p>Jmundo1: </p> <hr /> <div>Resources for machine translation between [[Lingala]] and [[Kikuyu]]. Code is stored in a [https://github.swarthmore.edu/apowell1/ling073-kik-lin GitHub Repository].<br /> ==Developed Resources==<br /> ===Parallel Corpus===<br /> <br /> A [https://github.swarthmore.edu/apowell1/ling073-lin-kik-corpus parallel corpus] created using two translations of Genesis is available on Github.<br /> <br /> ===Contrastive Grammar===<br /> Contrastive Grammar is detailed on the wiki page [[Lingala and Kikuyu/Contrastive Grammar]].<br /> <br /> ===Lexical Selection===<br /> Lexical selection is detailed on the wiki page [[Lingala and Kikuyu/Lexical Selection]].<br /> <br /> ===Structural Transfer===<br /> Structural transfer is detailed on the wiki page [[Lingala and Kikuyu/Structural Transfer]]<br /> <br /> ==Installing Unidecode For MT Evaluation==<br /> Most available Lingala text does not mark tones; however, the Lingala transducer has been configured using spellrelax to provide analyses and surface generation with tone markings included as accents on vowels. As a result, generated translations will include tone markings not present in the reference translation and artificially inflate evaluation metrics such as WER and PER. Fortunately, we can take advantage of the fact that accented characters are not present in ASCII to replace accented characters with their unaccented version on the command line.<br /> <br /> [https://pypi.python.org/pypi/Unidecode Unidecode] is a Python package that implements a function to convert Unicode strings into their closest ASCII approximation. For simple accented characters, this is precisely the mapping we need to remove the accents for evaluation against a reference translation. Unidecode can also be used on the command line with our machine translation system as follows: &lt;code&gt;echo /path/to/kik-corpus | apertium -d . kik-lin | unidecode&lt;/code&gt;. To install unidecode, it is recommended to use &lt;code&gt;pip&lt;/code&gt;, as experience suggests that &lt;code&gt;apt-get&lt;/code&gt; will not run the necessary installation scripts to make unidecode accessible in Bash.<br /> <br /> On the Ling 073 virtual machines, Pip can be installed using &lt;code&gt;sudo apt-get install python-pip&lt;/code&gt;. After installing pip, unidecode can be installed using &lt;code&gt; pip install unidecode&lt;/code&gt;. <br /> <br /> ==Lexical Selection==<br /> This section is part of the [[Lexical transfer]] assignment. In addition to adding many one-to-one mappings to our .dix file, we have included a couple of one-to-many mappings for each direction of translation. See [[Lingala and Kikuyu/Lexical Selection]] for further details.<br /> <br /> From Lingala to Kikuyu, we consider four one-to-many mappings:<br /> *The Lingala verb kokíma can be translated in English as both &quot;to run&quot; and &quot;to run away.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, hanyũk and ũr.<br /> *The Lingala verb kofánda can be translated in English as both &quot;to sit&quot; and &quot;to live&quot; (as in reside). Kikuyu has two distinct verbs corresponding to these English translations, respectively, ikar and tũũr.<br /> *The Lingala verb kobánza can be translated in English as both &quot;to think&quot; and &quot;to remember.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, ĩcir and ririkan.<br /> *The Lingala noun mabele can be translated in English as both &quot;sand&quot; and &quot;milk.&quot; Kikuyu has two distinct nouns corresponding to these English translations, respectively, thanga and ria.<br /> <br /> From Kikuyu to Lingala, we consider three one-to-many-mappings:<br /> *The Kikuyu verb igu can be translated in English as both &quot;to obey&quot; and &quot;to hear.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, kotosa and koyoka.<br /> *The Kikuyu verb rut can be translated in English as both &quot;to learn&quot; and &quot;to remove.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, koyekola and kokukula.<br /> *the Kikuyu noun ria can be translated in English as both &quot;milk&quot; and &quot;pond.&quot; Lingala has two distinct nouns corresponding to these English translations, respectively, mabele and etimá.<br /> <br /> ==Initial Evaluation==<br /> ===lin → kik evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 87.50%<br /> **PER: 87.50%<br /> **Number of words: 9<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 99.06%<br /> **PER: 94.34%<br /> **Number of words: 107<br /> **Number of unknown words: 71<br /> **Proportion of stems translated correctly:33.64%<br /> <br /> ===kik → lin evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 80%<br /> **PER: 80%<br /> **Number of words: 7 (This should really be 8, but the command we are supposed to use to count this outputs 7. The proportion remains unchanged in our case, however.)<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 98.13%<br /> **PER: 90.65%<br /> **Number of words: 106<br /> **Number of unknown words: 79 (Note: Using the commands on the wiki gives a slightly different count that using the apertium-eval-translator tool. We use the count given by the command, not the tool)<br /> **Proportion of stems translated correctly: 25.47%<br /> <br /> ==Final Evaluation==<br /> ===lin → kik Implementation Details===<br /> <br /> This direction of the pair was improved by improving the transducer to handle pronouns, proper names, and new types of morphology; improving the constraint grammar to better produce forms which agree properly; and adding transfer rules for several new input configurations including associatives and adjectives.<br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for demonstrative pronouns/determiners (proximal, distal, and anaphoric)<br /> **Example word: &quot;rĩrĩa&quot; (Class 5 singular distal demonstrative)<br /> **Old output: &lt;code&gt;^rĩrĩa/*rĩrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^rĩrĩa/pro&lt;rel&gt;&lt;cl5&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dist&gt;&lt;cl5&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for possessive pronouns/determiners<br /> **Example word: &quot;gĩakwa&quot; (Possessive pronoun for Class 7 singular object with first-person singular owner)<br /> **Old output: &lt;code&gt;^gĩakwa/*gĩakwa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gĩakwa/pro&lt;prn&gt;&lt;poss&gt;&lt;cl7&gt;&lt;sg&gt;&lt;sp1&gt;&lt;ssg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for relative pronouns<br /> **Example word: ũrĩa (Class 1 singular relative pronoun, as well as other noun classes and other types of pronouns)<br /> **Old output: &lt;code&gt;^ũrĩa/*ũrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^ũrĩa/pro&lt;prn&gt;&lt;rel&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl14&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl14&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for personal pronouns<br /> **Example word: yo (Class 3 plural personal pronoun)<br /> **Old output: &lt;code&gt;^yo/*yo$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yo/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;sg&gt;$&lt;/code&gt;<br /> <br /> *Support for proper names (proper name tag introduced, and several names added)<br /> **Example word: Kenya (Kenya)<br /> **Old output: &lt;code&gt;^Kenya/Kenya$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kenya/Kenya&lt;n&gt;&lt;prop&gt;$&lt;/code&gt;<br /> <br /> *Support for Class 16 and 17 nouns (which tend to have to do with place), and their agreement on adjectives<br /> **Example: kũndũ guothe (all places)<br /> **Old output: &lt;code&gt;^kũndũ/*kũndũ$ ^guothe/*guothe$&lt;/code&gt;<br /> **New output: &lt;code&gt;^kũndũ/ndũ&lt;n&gt;&lt;cl17&gt;&lt;sg&gt;&lt;indef&gt;$ ^guothe/othe&lt;adj&gt;&lt;cl17&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for unique reversive middle voice morphology<br /> **Example word: rũthiũrũrũkĩirie (surround, roughly)<br /> **Old output: &lt;code&gt;^makĩinũkae/*makĩinũka$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^makĩinũka/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for infinitives<br /> **Example word: gũthambĩra (to swim)<br /> **Old output: &lt;code&gt;^gũthambĩra/*gũthambĩra$ ^guothe/*guothe$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gũthambĩra/^gũthambĩra/thambĩr&lt;v&gt;&lt;inf&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *An attempt was made to implement object agreement morphology on verbs. Such an attempt would necessitate adding twoc rules for each noun class, and doing so made twoc take far too long to compile. This attempt was therefore abandoned, but if a workaround were to be found for twoc, the twoc rules and lexc code are commented out in their respective files.<br /> <br /> ====Additional Smoothing of Transducer====<br /> <br /> *Support for ĩ to y alternation in certain cases (implemented through twol)<br /> **Example word: yarĩ (copula form)<br /> **Old output: &lt;code&gt;^yarĩ/*yarĩ$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yarĩ/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p2&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p13&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl7&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p10&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for ũ to u alternation before o<br /> **Example word: kuona (to see)<br /> **Old output: &lt;code&gt;^kuona/*kuona$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^kuona/on&lt;v&gt;&lt;inf&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> ====Additional Constraint Grammar Rules====<br /> <br /> Note: several of the rules in this section required multiple rlx rules to implement, because they act on various different aspects of agreement, including number, noun class, and person.<br /> <br /> *If I am a verb, demonstrative, or associative inflected to agree with a certain number/noun class, select me if I am preceded by an adjective preceded by a noun of that number/noun class.<br /> **Example: maaĩ maingĩ me (many waters are)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;Maaĩ&gt;&quot;<br /> &quot;aĩ&quot; n cl5 pl<br /> &quot;&lt;maingĩ&gt;&quot;<br /> &quot;ingĩ&quot; adj cl5 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl9p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl14 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl11p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl15 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl12p6 pl SELECT:27<br /> &quot;&lt;me&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:33<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/noun class, select me if I am preceded by a demonstrative preceded by a noun of that number/person/noun class.<br /> **Example: andũ acio nĩmagakorwo (those people will find)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;andũ&gt;&quot;<br /> &quot;ndũ&quot; n cl1 pl<br /> &quot;&lt;acio&gt;&quot;<br /> &quot;pro&quot; prn dem ana cl1 pl<br /> &quot;&lt;nĩmagakorwo&gt;&quot;<br /> &quot;kor&quot; v pres pass p3 cl1 pl foc SELECT:40<br /> &quot;kor&quot; v remfut pass p3 cl1 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl12p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl12p6 pl foc SELECT:40<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/person/noun class, select me if I am preceded by a number preceded by an associative preceded by a noun of that noun class.<br /> **Example: nyamũ ya mũthia nĩ (the last animal is)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;nyamũ&gt;&quot;<br /> &quot;nyamũ&quot; n cl9p10 pl<br /> &quot;nyamũ&quot; n cl9p10 sg<br /> &quot;&lt;ya&gt;&quot;<br /> &quot;a&quot; assoc cl9p10 sg SELECT:27<br /> ; &quot;a&quot; assoc cl3 pl SELECT:27<br /> ; &quot;a&quot; assoc cl9p6 sg SELECT:27<br /> &quot;&lt;mũthia&gt;&quot;<br /> &quot;mũthia&quot; num<br /> &quot;&lt;nĩ&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl9p10 pl SELECT:46<br /> &quot;rĩ&quot; cop pres p3 cl9p10 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 sg SELECT:46<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a preposition followed by a verb, remove me.<br /> **Example: na ndĩarĩ na kĩndũ (and was of nothing)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> ; &quot;na&quot; pr REMOVE:68<br /> &quot;&lt;ndĩarĩ&gt;&quot;<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p6 sg<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p10 sg<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;kĩndũ&gt;&quot;<br /> &quot;ndũ&quot; n cl7 sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> Note: None of the examples here will work, because there remains a bug in generation in the Kikuyu transducer. It cannot generate most forms it can analyze. All the forms outputted in these examples would be generated correctly if this bug is fixed, without other changes to the Kikuyu transducer.<br /> <br /> The chunker is the only part of the structural transfer system which has anything beyond the default, so only the chunker and final output is given.<br /> *Implemented transfer of adjectives<br /> **Example: ezibele monene<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^rango&lt;n&gt;&lt;cl3&gt;&lt;sg&gt;$ ^nene&lt;adj&gt;&lt;cl3&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}&lt;/code&gt;<br /> **Final output: &lt;code&gt; #rango #nene &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũrango mũnene &lt;/code&gt;<br /> <br /> *Implemented transfer of imperatives<br /> **Example: koyemba<br /> **Output of chunker: &lt;code&gt;^imperative&lt;verb&gt;&lt;imperative&gt;{^in&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output: &lt;code&gt;#in&lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed:&lt;code&gt;ina&lt;/code&gt;<br /> *Implemented transfer of associative<br /> **Example: bato ya mokili (the people of the world)<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;pl&gt;$ ^a&lt;assoc&gt;&lt;cl1&gt;&lt;pl&gt;$ ^thĩ&lt;n&gt;&lt;cl9p10&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ a thĩ &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; andũ a thĩ &lt;/code&gt;<br /> <br /> *Implemented transfer of demonstratives<br /> **Example: moto óyo (this person)<br /> **Output of chunker: &lt;code&gt;^nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;sg&gt;$ ^pro&lt;prn&gt;&lt;dem&gt;&lt;prox&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ ũyũ&lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũndũ ũyũ &lt;/code&gt;<br /> <br /> ===kik → lin Implementation Details===<br /> This direction of the RBMT system was improved by adding items to the bilingual dictionary, adding additional morphology to the transducer, writing more constraint grammar rules, and writing additional transfer rules. Specific implementation details with examples of improved output are detailed in the sections that follow.<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for proper nouns<br /> **Example sentence: &quot;Nzambe&quot; (God)<br /> **Old output: &lt;code&gt;^Nzambe/*Nzambe$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nzambe/Nzambe&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for imperative mood<br /> **Example sentence: &quot;Kíma!&quot; (Run!)<br /> **Old output: &lt;code&gt;^Kíma/*Kíma$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kíma/kokíma&lt;v&gt;&lt;imp&gt;$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for intensifying suffixes (-a becomes -áká to denote certainty (infinitive, future) or insistence (imperative))<br /> **Example sentence 1: &quot;Azalákí kopésáká&quot; (He was really distributing...)<br /> **Old output: &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/*kopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/kopésa&lt;v&gt;&lt;inf&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 2: &quot;Akopésáká&quot; (He will really give...)<br /> **Old output: &lt;code&gt;^Akopésáká/*Akopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Akopésáká/kopésa&lt;v&gt;&lt;intns&gt;&lt;fut&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 3: &quot;Pésáká&quot; (Now do give...)<br /> **Old output: &lt;code&gt;^Pésáká/*Pésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Pésáká/kopésa&lt;v&gt;&lt;imp&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for more than 1 radical extension<br /> **Example sentence: &quot;Nabángisamákí na Pierre&quot; (I was scared by Pierre, or more literally, I was made, by Pierre, to fear something/someone)<br /> **Old output: &lt;code&gt;^Nabángisamákí/*Nabángisamákí$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nabángisamákí/kobánga&lt;v&gt;&lt;caus&gt;&lt;pass&gt;&lt;urp&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for numerals<br /> **Example sentence: &quot;zómi&quot; (ten)<br /> **Old output: &lt;code&gt;^zómi/*zómi$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^zómi/zómi&lt;num&gt;&lt;ord&gt;/zómi&lt;num&gt;&lt;card&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for quantifiers<br /> **Example sentence: &quot;mbelí nyónso&quot; (each knife)<br /> **Old output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/*nyónso$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/nyónso&lt;det&gt;&lt;qnt&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> ====Additional Constraint Grammar Rules====<br /> *If I am an adjective and a quantifier but I do not agree in number with the preceding noun, I must be a quantifier<br /> **Example sentence: &quot;bilóko moké&quot; (A few things, &#039;&#039;not&#039;&#039; small things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> &quot;moké&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> ; &quot;moké&quot; adj sg REMOVE:21<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a singular adjective, I cannot be plural<br /> **Example sentence: &quot;nzelá molaí&quot; (a long road)<br /> **Old output (below) <br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> ; &quot;nzelá&quot; n cl9 pl REMOVE:25<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a plural adjective, I cannot be singular<br /> **Example sentence: &quot;nzelá milaí&quot; (long roads)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 pl<br /> ; &quot;nzelá&quot; n cl9 sg REMOVE:28<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be an ordinal and a cardinal and am modifying a plural noun, I am not an ordinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;bilóko míbalé&quot; (two things, &#039;&#039;not&#039;&#039; the second thing)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num card<br /> ; &quot;míbalé&quot; num ord REMOVE:31<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be and ordinal and a cardinal and am modifying a a singular noun, I am not a cardinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;elóko mibalé&quot; (the second thing, &#039;&#039;not&#039;&#039; two things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> ; &quot;míbalé&quot; num card REMOVE:34<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be a demonstrative nominal and a relative pronoun and am proceeded by a noun, I cannot be a demonstrative nominal.<br /> **Example sentence: &quot;Mobali óyo amonaki ngáí&quot; (The man who saw me...)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> &quot;óyo&quot; prn dem sg<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> ; &quot;óyo&quot; prn dem sg REMOVE:37<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> *Enhance existing basic verb transfer rule to work with newly implemented imperative<br /> **Example sentence: &quot;ina&quot; (sing)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#koyemba&lt;/code&gt;<br /> **New output<br /> ***Tagger:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;imp&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;yemba&lt;/code&gt;<br /> <br /> *Map Kikuyu associative to Lingala connective by removing extra agreement tags found in Kikuyu<br /> **Example sentence: &quot;wa&quot;<br /> **Old output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#ya&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;ya&lt;/code&gt;<br /> <br /> *Basic adjective transfer: keep number markings if a variable adjective in Lingala, discard otherwise<br /> **Example sentence 1: anene (big)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#monéne&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;minéne&lt;/code&gt;<br /> **Example sentence 2: mata (sour)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#bololo&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;bololo&lt;/code&gt;<br /> <br /> ===lin Transducer Evaluation===<br /> *Precision (annotated corpus): 80.53%<br /> *Recall (annotated corpus): 55.64%<br /> *Coverage (large corpus): 62.10%<br /> *Number of words (large corpus): 275,546 (actual), 304,425 (tokenized)<br /> *Number of stems in transducer: 273<br /> <br /> ===kik Transducer Evaluation===<br /> *Precision (annotated corpus): 79.24%<br /> *Recall (annotated corpus): 91.52%<br /> *Coverage (large corpus): 58.39%<br /> *Number of words (large corpus): 107564<br /> *Number of stems in transducer: about 267 (nouns, verbs, adjectives, and adverbs)<br /> <br /> ===lin → kik MT Evaluation===<br /> Note that the bug in the Kikuyu generator significantly alters these numbers. They should improve significantly if the bug is resolved.<br /> *WER (longer corpus): 96.49%<br /> *PER (longer corpus): 92.98%<br /> *Percentage unknown words: 13.91%<br /> *Percentage known words: 86.09%<br /> *Trimmed coverage over longer corpus [lin-kik.automorf.bin applied to lin.longer.txt]: 89.29%<br /> *Trimmed coverage over large corpus [lin-kik.automorf.bin applied to lin.corpus.large.txt]: 48.90%<br /> *Number of stems in longer corpus: 117 (actual) 140 (tokenized, perhaps including punctuation)<br /> *Number of stems in large corpus: 275,546 (actual), 304,425 (tokenized)<br /> <br /> ===kik → lin MT Evaluation===<br /> *WER (longer corpus): 86.32%<br /> *PER (longer corpus): 75.21%<br /> *Percentage unknown words: 39.50% [Actually probably a bit higher since some words were marked with an &quot;@&quot; and not translated while the script only looks for words marked with a &quot;*&quot;]<br /> *Percentage known words: 60.60%<br /> *Trimmed coverage over longer corpus [kik-lin.automorf.bin applied to kik.corpus.larger.txt]: 71.43%<br /> *Trimmed coverage over large corpus [kik-lin.automorf.bin applied to kik.longer.txt]: 46.40%<br /> *Number of stems in longer corpus: 115<br /> *Number of stems in large corpus: 107564<br /> <br /> <br /> [[Category:Sp17_TranslationPairs]]<br /> [[Category:Lingala]]<br /> [[Category:Kikuyu]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Lingala_and_Kikuyu&diff=4589 2017-04-18T02:23:15Z

<p>Jmundo1: </p> <hr /> <div>Resources for machine translation between [[Lingala]] and [[Kikuyu]]. Code is stored in a [https://github.swarthmore.edu/apowell1/ling073-kik-lin GitHub Repository].<br /> ==Developed Resources==<br /> ===Parallel Corpus===<br /> <br /> A [https://github.swarthmore.edu/apowell1/ling073-lin-kik-corpus parallel corpus] created using two translations of Genesis is available on Github.<br /> <br /> ===Contrastive Grammar===<br /> Contrastive Grammar is detailed on the wiki page [[Lingala and Kikuyu/Contrastive Grammar]].<br /> <br /> ===Lexical Selection===<br /> Lexical selection is detailed on the wiki page [[Lingala and Kikuyu/Lexical Selection]].<br /> <br /> ===Structural Transfer===<br /> Structural transfer is detailed on the wiki page [[Lingala and Kikuyu/Structural Transfer]]<br /> <br /> ==Installing Unidecode For MT Evaluation==<br /> Most available Lingala text does not mark tones; however, the Lingala transducer has been configured using spellrelax to provide analyses and surface generation with tone markings included as accents on vowels. As a result, generated translations will include tone markings not present in the reference translation and artificially inflate evaluation metrics such as WER and PER. Fortunately, we can take advantage of the fact that accented characters are not present in ASCII to replace accented characters with their unaccented version on the command line.<br /> <br /> [https://pypi.python.org/pypi/Unidecode Unidecode] is a Python package that implements a function to convert Unicode strings into their closest ASCII approximation. For simple accented characters, this is precisely the mapping we need to remove the accents for evaluation against a reference translation. Unidecode can also be used on the command line with our machine translation system as follows: &lt;code&gt;echo /path/to/kik-corpus | apertium -d . kik-lin | unidecode&lt;/code&gt;. To install unidecode, it is recommended to use &lt;code&gt;pip&lt;/code&gt;, as experience suggests that &lt;code&gt;apt-get&lt;/code&gt; will not run the necessary installation scripts to make unidecode accessible in Bash.<br /> <br /> On the Ling 073 virtual machines, Pip can be installed using &lt;code&gt;sudo apt-get install python-pip&lt;/code&gt;. After installing pip, unidecode can be installed using &lt;code&gt; pip install unidecode&lt;/code&gt;. <br /> <br /> ==Lexical Selection==<br /> This section is part of the [[Lexical transfer]] assignment. In addition to adding many one-to-one mappings to our .dix file, we have included a couple of one-to-many mappings for each direction of translation. See [[Lingala and Kikuyu/Lexical Selection]] for further details.<br /> <br /> From Lingala to Kikuyu, we consider four one-to-many mappings:<br /> *The Lingala verb kokíma can be translated in English as both &quot;to run&quot; and &quot;to run away.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, hanyũk and ũr.<br /> *The Lingala verb kofánda can be translated in English as both &quot;to sit&quot; and &quot;to live&quot; (as in reside). Kikuyu has two distinct verbs corresponding to these English translations, respectively, ikar and tũũr.<br /> *The Lingala verb kobánza can be translated in English as both &quot;to think&quot; and &quot;to remember.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, ĩcir and ririkan.<br /> *The Lingala noun mabele can be translated in English as both &quot;sand&quot; and &quot;milk.&quot; Kikuyu has two distinct nouns corresponding to these English translations, respectively, thanga and ria.<br /> <br /> From Kikuyu to Lingala, we consider three one-to-many-mappings:<br /> *The Kikuyu verb igu can be translated in English as both &quot;to obey&quot; and &quot;to hear.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, kotosa and koyoka.<br /> *The Kikuyu verb rut can be translated in English as both &quot;to learn&quot; and &quot;to remove.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, koyekola and kokukula.<br /> *the Kikuyu noun ria can be translated in English as both &quot;milk&quot; and &quot;pond.&quot; Lingala has two distinct nouns corresponding to these English translations, respectively, mabele and etimá.<br /> <br /> ==Initial Evaluation==<br /> ===lin → kik evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 87.50%<br /> **PER: 87.50%<br /> **Number of words: 9<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 99.06%<br /> **PER: 94.34%<br /> **Number of words: 107<br /> **Number of unknown words: 71<br /> **Proportion of stems translated correctly:33.64%<br /> <br /> ===kik → lin evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 80%<br /> **PER: 80%<br /> **Number of words: 7 (This should really be 8, but the command we are supposed to use to count this outputs 7. The proportion remains unchanged in our case, however.)<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 98.13%<br /> **PER: 90.65%<br /> **Number of words: 106<br /> **Number of unknown words: 79 (Note: Using the commands on the wiki gives a slightly different count that using the apertium-eval-translator tool. We use the count given by the command, not the tool)<br /> **Proportion of stems translated correctly: 25.47%<br /> <br /> ==Final Evaluation==<br /> ===lin → kik Implementation Details===<br /> <br /> This direction of the pair was improved by improving the transducer to handle pronouns, proper names, and new types of morphology; improving the constraint grammar to better produce forms which agree properly; and adding transfer rules for several new input configurations including associatives and adjectives.<br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for demonstrative pronouns/determiners (proximal, distal, and anaphoric)<br /> **Example word: &quot;rĩrĩa&quot; (Class 5 singular distal demonstrative)<br /> **Old output: &lt;code&gt;^rĩrĩa/*rĩrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^rĩrĩa/pro&lt;rel&gt;&lt;cl5&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dist&gt;&lt;cl5&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for possessive pronouns/determiners<br /> **Example word: &quot;gĩakwa&quot; (Possessive pronoun for Class 7 singular object with first-person singular owner)<br /> **Old output: &lt;code&gt;^gĩakwa/*gĩakwa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gĩakwa/pro&lt;prn&gt;&lt;poss&gt;&lt;cl7&gt;&lt;sg&gt;&lt;sp1&gt;&lt;ssg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for relative pronouns<br /> **Example word: ũrĩa (Class 1 singular relative pronoun, as well as other noun classes and other types of pronouns)<br /> **Old output: &lt;code&gt;^ũrĩa/*ũrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^ũrĩa/pro&lt;prn&gt;&lt;rel&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl14&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl14&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for personal pronouns<br /> **Example word: yo (Class 3 plural personal pronoun)<br /> **Old output: &lt;code&gt;^yo/*yo$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yo/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;sg&gt;$&lt;/code&gt;<br /> <br /> *Support for proper names (proper name tag introduced, and several names added)<br /> **Example word: Kenya (Kenya)<br /> **Old output: &lt;code&gt;^Kenya/Kenya$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kenya/Kenya&lt;n&gt;&lt;prop&gt;$&lt;/code&gt;<br /> <br /> *Support for Class 16 and 17 nouns (which tend to have to do with place), and their agreement on adjectives<br /> **Example: kũndũ guothe (all places)<br /> **Old output: &lt;code&gt;^kũndũ/*kũndũ$ ^guothe/*guothe$&lt;/code&gt;<br /> **New output: &lt;code&gt;^kũndũ/ndũ&lt;n&gt;&lt;cl17&gt;&lt;sg&gt;&lt;indef&gt;$ ^guothe/othe&lt;adj&gt;&lt;cl17&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for unique reversive middle voice morphology<br /> **Example word: rũthiũrũrũkĩirie (surround, roughly)<br /> **Old output: &lt;code&gt;^makĩinũkae/*makĩinũka$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^makĩinũka/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for infinitives<br /> **Example word: gũthambĩra (to swim)<br /> **Old output: &lt;code&gt;^gũthambĩra/*gũthambĩra$ ^guothe/*guothe$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gũthambĩra/^gũthambĩra/thambĩr&lt;v&gt;&lt;inf&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *An attempt was made to implement object agreement morphology on verbs. Such an attempt would necessitate adding twoc rules for each noun class, and doing so made twoc take far too long to compile. This attempt was therefore abandoned, but if a workaround were to be found for twoc, the twoc rules and lexc code are commented out in their respective files.<br /> <br /> ====Additional Smoothing of Transducer====<br /> <br /> *Support for ĩ to y alternation in certain cases (implemented through twol)<br /> **Example word: yarĩ (copula form)<br /> **Old output: &lt;code&gt;^yarĩ/*yarĩ$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yarĩ/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p2&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p13&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl7&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p10&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for ũ to u alternation before o<br /> **Example word: kuona (to see)<br /> **Old output: &lt;code&gt;^kuona/*kuona$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^kuona/on&lt;v&gt;&lt;inf&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> ====Additional Constraint Grammar Rules====<br /> <br /> Note: several of the rules in this section required multiple rlx rules to implement, because they act on various different aspects of agreement, including number, noun class, and person.<br /> <br /> *If I am a verb, demonstrative, or associative inflected to agree with a certain number/noun class, select me if I am preceded by an adjective preceded by a noun of that number/noun class.<br /> **Example: maaĩ maingĩ me (many waters are)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;Maaĩ&gt;&quot;<br /> &quot;aĩ&quot; n cl5 pl<br /> &quot;&lt;maingĩ&gt;&quot;<br /> &quot;ingĩ&quot; adj cl5 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl9p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl14 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl11p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl15 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl12p6 pl SELECT:27<br /> &quot;&lt;me&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:33<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/noun class, select me if I am preceded by a demonstrative preceded by a noun of that number/person/noun class.<br /> **Example: andũ acio nĩmagakorwo (those people will find)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;andũ&gt;&quot;<br /> &quot;ndũ&quot; n cl1 pl<br /> &quot;&lt;acio&gt;&quot;<br /> &quot;pro&quot; prn dem ana cl1 pl<br /> &quot;&lt;nĩmagakorwo&gt;&quot;<br /> &quot;kor&quot; v pres pass p3 cl1 pl foc SELECT:40<br /> &quot;kor&quot; v remfut pass p3 cl1 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl12p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl12p6 pl foc SELECT:40<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/person/noun class, select me if I am preceded by a number preceded by an associative preceded by a noun of that noun class.<br /> **Example: nyamũ ya mũthia nĩ (the last animal is)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;nyamũ&gt;&quot;<br /> &quot;nyamũ&quot; n cl9p10 pl<br /> &quot;nyamũ&quot; n cl9p10 sg<br /> &quot;&lt;ya&gt;&quot;<br /> &quot;a&quot; assoc cl9p10 sg SELECT:27<br /> ; &quot;a&quot; assoc cl3 pl SELECT:27<br /> ; &quot;a&quot; assoc cl9p6 sg SELECT:27<br /> &quot;&lt;mũthia&gt;&quot;<br /> &quot;mũthia&quot; num<br /> &quot;&lt;nĩ&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl9p10 pl SELECT:46<br /> &quot;rĩ&quot; cop pres p3 cl9p10 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 sg SELECT:46<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a preposition followed by a verb, remove me.<br /> **Example: na ndĩarĩ na kĩndũ (and was of nothing)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> ; &quot;na&quot; pr REMOVE:68<br /> &quot;&lt;ndĩarĩ&gt;&quot;<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p6 sg<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p10 sg<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;kĩndũ&gt;&quot;<br /> &quot;ndũ&quot; n cl7 sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> Note: None of the examples here will work, because there remains a bug in generation in the Kikuyu transducer. It cannot generate most forms it can analyze. All the forms outputted in these examples would be generated correctly if this bug is fixed, without other changes to the Kikuyu transducer.<br /> <br /> The chunker is the only part of the structural transfer system which has anything beyond the default, so only the chunker and final output is given.<br /> *Implemented transfer of adjectives<br /> **Example: ezibele monene<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^rango&lt;n&gt;&lt;cl3&gt;&lt;sg&gt;$ ^nene&lt;adj&gt;&lt;cl3&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}&lt;/code&gt;<br /> **Final output: &lt;code&gt; #rango #nene &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũrango mũnene &lt;/code&gt;<br /> <br /> *Implemented transfer of imperatives<br /> **Example: koyemba<br /> **Output of chunker: &lt;code&gt;^imperative&lt;verb&gt;&lt;imperative&gt;{^in&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output: &lt;code&gt;#in&lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed:&lt;code&gt;ina&lt;/code&gt;<br /> *Implemented transfer of associative<br /> **Example: bato ya mokili (the people of the world)<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;pl&gt;$ ^a&lt;assoc&gt;&lt;cl1&gt;&lt;pl&gt;$ ^thĩ&lt;n&gt;&lt;cl9p10&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ a thĩ &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; andũ a thĩ &lt;/code&gt;<br /> <br /> *Implemented transfer of demonstratives<br /> **Example: moto óyo (this person)<br /> **Output of chunker: &lt;code&gt;^nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;sg&gt;$ ^pro&lt;prn&gt;&lt;dem&gt;&lt;prox&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ ũyũ&lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũndũ ũyũ &lt;/code&gt;<br /> <br /> ===kik → lin Implementation Details===<br /> This direction of the RBMT system was improved by adding items to the bilingual dictionary, adding additional morphology to the transducer, writing more constraint grammar rules, and writing additional transfer rules. Specific implementation details with examples of improved output are detailed in the sections that follow.<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for proper nouns<br /> **Example sentence: &quot;Nzambe&quot; (God)<br /> **Old output: &lt;code&gt;^Nzambe/*Nzambe$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nzambe/Nzambe&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for imperative mood<br /> **Example sentence: &quot;Kíma!&quot; (Run!)<br /> **Old output: &lt;code&gt;^Kíma/*Kíma$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kíma/kokíma&lt;v&gt;&lt;imp&gt;$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for intensifying suffixes (-a becomes -áká to denote certainty (infinitive, future) or insistence (imperative))<br /> **Example sentence 1: &quot;Azalákí kopésáká&quot; (He was really distributing...)<br /> **Old output: &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/*kopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/kopésa&lt;v&gt;&lt;inf&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 2: &quot;Akopésáká&quot; (He will really give...)<br /> **Old output: &lt;code&gt;^Akopésáká/*Akopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Akopésáká/kopésa&lt;v&gt;&lt;intns&gt;&lt;fut&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 3: &quot;Pésáká&quot; (Now do give...)<br /> **Old output: &lt;code&gt;^Pésáká/*Pésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Pésáká/kopésa&lt;v&gt;&lt;imp&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for more than 1 radical extension<br /> **Example sentence: &quot;Nabángisamákí na Pierre&quot; (I was scared by Pierre, or more literally, I was made, by Pierre, to fear something/someone)<br /> **Old output: &lt;code&gt;^Nabángisamákí/*Nabángisamákí$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nabángisamákí/kobánga&lt;v&gt;&lt;caus&gt;&lt;pass&gt;&lt;urp&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for numerals<br /> **Example sentence: &quot;zómi&quot; (ten)<br /> **Old output: &lt;code&gt;^zómi/*zómi$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^zómi/zómi&lt;num&gt;&lt;ord&gt;/zómi&lt;num&gt;&lt;card&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for quantifiers<br /> **Example sentence: &quot;mbelí nyónso&quot; (each knife)<br /> **Old output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/*nyónso$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/nyónso&lt;det&gt;&lt;qnt&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> ====Additional Constraint Grammar Rules====<br /> *If I am an adjective and a quantifier but I do not agree in number with the preceding noun, I must be a quantifier<br /> **Example sentence: &quot;bilóko moké&quot; (A few things, &#039;&#039;not&#039;&#039; small things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> &quot;moké&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> ; &quot;moké&quot; adj sg REMOVE:21<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a singular adjective, I cannot be plural<br /> **Example sentence: &quot;nzelá molaí&quot; (a long road)<br /> **Old output (below) <br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> ; &quot;nzelá&quot; n cl9 pl REMOVE:25<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a plural adjective, I cannot be singular<br /> **Example sentence: &quot;nzelá milaí&quot; (long roads)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 pl<br /> ; &quot;nzelá&quot; n cl9 sg REMOVE:28<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be an ordinal and a cardinal and am modifying a plural noun, I am not an ordinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;bilóko míbalé&quot; (two things, &#039;&#039;not&#039;&#039; the second thing)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num card<br /> ; &quot;míbalé&quot; num ord REMOVE:31<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be and ordinal and a cardinal and am modifying a a singular noun, I am not a cardinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;elóko mibalé&quot; (the second thing, &#039;&#039;not&#039;&#039; two things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> ; &quot;míbalé&quot; num card REMOVE:34<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be a demonstrative nominal and a relative pronoun and am proceeded by a noun, I cannot be a demonstrative nominal.<br /> **Example sentence: &quot;Mobali óyo amonaki ngáí&quot; (The man who saw me...)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> &quot;óyo&quot; prn dem sg<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> ; &quot;óyo&quot; prn dem sg REMOVE:37<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> *Enhance existing basic verb transfer rule to work with newly implemented imperative<br /> **Example sentence: &quot;ina&quot; (sing)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#koyemba&lt;/code&gt;<br /> **New output<br /> ***Tagger:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;imp&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;yemba&lt;/code&gt;<br /> <br /> *Map Kikuyu associative to Lingala connective by removing extra agreement tags found in Kikuyu<br /> **Example sentence: &quot;wa&quot;<br /> **Old output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#ya&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;ya&lt;/code&gt;<br /> <br /> *Basic adjective transfer: keep number markings if a variable adjective in Lingala, discard otherwise<br /> **Example sentence 1: anene (big)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#monéne&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;minéne&lt;/code&gt;<br /> **Example sentence 2: mata (sour)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#bololo&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;bololo&lt;/code&gt;<br /> <br /> ===lin Transducer Evaluation===<br /> *Precision (annotated corpus): 80.53%<br /> *Recall (annotated corpus): 55.64%<br /> *Coverage (large corpus): 62.10%<br /> *Number of words (large corpus): 275,546 (actual), 304,425 (tokenized)<br /> *Number of stems in transducer: 273<br /> <br /> ===kik Transducer Evaluation===<br /> *Precision (annotated corpus):<br /> *Recall (annotated corpus): <br /> *Coverage (large corpus): 58.39%<br /> *Number of words (large corpus): 107564<br /> *Number of stems in transducer: about 267 (nouns, verbs, adjectives, and adverbs)<br /> <br /> ===lin → kik MT Evaluation===<br /> Note that the bug in the Kikuyu generator significantly alters these numbers. They should improve significantly if the bug is resolved.<br /> *WER (longer corpus): 96.49%<br /> *PER (longer corpus): 92.98%<br /> *Percentage unknown words: 13.91%<br /> *Percentage known words: 86.09%<br /> *Trimmed coverage over longer corpus [lin-kik.automorf.bin applied to lin.longer.txt]: 89.29%<br /> *Trimmed coverage over large corpus [lin-kik.automorf.bin applied to lin.corpus.large.txt]: 48.90%<br /> *Number of stems in longer corpus: 117 (actual) 140 (tokenized, perhaps including punctuation)<br /> *Number of stems in large corpus: 275,546 (actual), 304,425 (tokenized)<br /> <br /> ===kik → lin MT Evaluation===<br /> *WER (longer corpus): 86.32%<br /> *PER (longer corpus): 75.21%<br /> *Percentage unknown words: 39.50% [Actually probably a bit higher since some words were marked with an &quot;@&quot; and not translated while the script only looks for words marked with a &quot;*&quot;]<br /> *Percentage known words: 60.60%<br /> *Trimmed coverage over longer corpus [kik-lin.automorf.bin applied to kik.corpus.larger.txt]: 71.43%<br /> *Trimmed coverage over large corpus [kik-lin.automorf.bin applied to kik.longer.txt]: 46.40%<br /> *Number of stems in longer corpus: 115<br /> *Number of stems in large corpus: 107564<br /> <br /> <br /> [[Category:Sp17_TranslationPairs]]<br /> [[Category:Lingala]]<br /> [[Category:Kikuyu]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Lingala_and_Kikuyu&diff=4588 2017-04-18T02:20:41Z

<p>Jmundo1: </p> <hr /> <div>Resources for machine translation between [[Lingala]] and [[Kikuyu]]. Code is stored in a [https://github.swarthmore.edu/apowell1/ling073-kik-lin GitHub Repository].<br /> ==Developed Resources==<br /> ===Parallel Corpus===<br /> <br /> A [https://github.swarthmore.edu/apowell1/ling073-lin-kik-corpus parallel corpus] created using two translations of Genesis is available on Github.<br /> <br /> ===Contrastive Grammar===<br /> Contrastive Grammar is detailed on the wiki page [[Lingala and Kikuyu/Contrastive Grammar]].<br /> <br /> ===Lexical Selection===<br /> Lexical selection is detailed on the wiki page [[Lingala and Kikuyu/Lexical Selection]].<br /> <br /> ===Structural Transfer===<br /> Structural transfer is detailed on the wiki page [[Lingala and Kikuyu/Structural Transfer]]<br /> <br /> ==Installing Unidecode For MT Evaluation==<br /> Most available Lingala text does not mark tones; however, the Lingala transducer has been configured using spellrelax to provide analyses and surface generation with tone markings included as accents on vowels. As a result, generated translations will include tone markings not present in the reference translation and artificially inflate evaluation metrics such as WER and PER. Fortunately, we can take advantage of the fact that accented characters are not present in ASCII to replace accented characters with their unaccented version on the command line.<br /> <br /> [https://pypi.python.org/pypi/Unidecode Unidecode] is a Python package that implements a function to convert Unicode strings into their closest ASCII approximation. For simple accented characters, this is precisely the mapping we need to remove the accents for evaluation against a reference translation. Unidecode can also be used on the command line with our machine translation system as follows: &lt;code&gt;echo /path/to/kik-corpus | apertium -d . kik-lin | unidecode&lt;/code&gt;. To install unidecode, it is recommended to use &lt;code&gt;pip&lt;/code&gt;, as experience suggests that &lt;code&gt;apt-get&lt;/code&gt; will not run the necessary installation scripts to make unidecode accessible in Bash.<br /> <br /> On the Ling 073 virtual machines, Pip can be installed using &lt;code&gt;sudo apt-get install python-pip&lt;/code&gt;. After installing pip, unidecode can be installed using &lt;code&gt; pip install unidecode&lt;/code&gt;. <br /> <br /> ==Lexical Selection==<br /> This section is part of the [[Lexical transfer]] assignment. In addition to adding many one-to-one mappings to our .dix file, we have included a couple of one-to-many mappings for each direction of translation. See [[Lingala and Kikuyu/Lexical Selection]] for further details.<br /> <br /> From Lingala to Kikuyu, we consider four one-to-many mappings:<br /> *The Lingala verb kokíma can be translated in English as both &quot;to run&quot; and &quot;to run away.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, hanyũk and ũr.<br /> *The Lingala verb kofánda can be translated in English as both &quot;to sit&quot; and &quot;to live&quot; (as in reside). Kikuyu has two distinct verbs corresponding to these English translations, respectively, ikar and tũũr.<br /> *The Lingala verb kobánza can be translated in English as both &quot;to think&quot; and &quot;to remember.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, ĩcir and ririkan.<br /> *The Lingala noun mabele can be translated in English as both &quot;sand&quot; and &quot;milk.&quot; Kikuyu has two distinct nouns corresponding to these English translations, respectively, thanga and ria.<br /> <br /> From Kikuyu to Lingala, we consider three one-to-many-mappings:<br /> *The Kikuyu verb igu can be translated in English as both &quot;to obey&quot; and &quot;to hear.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, kotosa and koyoka.<br /> *The Kikuyu verb rut can be translated in English as both &quot;to learn&quot; and &quot;to remove.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, koyekola and kokukula.<br /> *the Kikuyu noun ria can be translated in English as both &quot;milk&quot; and &quot;pond.&quot; Lingala has two distinct nouns corresponding to these English translations, respectively, mabele and etimá.<br /> <br /> ==Initial Evaluation==<br /> ===lin → kik evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 87.50%<br /> **PER: 87.50%<br /> **Number of words: 9<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 99.06%<br /> **PER: 94.34%<br /> **Number of words: 107<br /> **Number of unknown words: 71<br /> **Proportion of stems translated correctly:33.64%<br /> <br /> ===kik → lin evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 80%<br /> **PER: 80%<br /> **Number of words: 7 (This should really be 8, but the command we are supposed to use to count this outputs 7. The proportion remains unchanged in our case, however.)<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 98.13%<br /> **PER: 90.65%<br /> **Number of words: 106<br /> **Number of unknown words: 79 (Note: Using the commands on the wiki gives a slightly different count that using the apertium-eval-translator tool. We use the count given by the command, not the tool)<br /> **Proportion of stems translated correctly: 25.47%<br /> <br /> ==Final Evaluation==<br /> ===lin → kik Implementation Details===<br /> <br /> This direction of the pair was improved by improving the transducer to handle pronouns, proper names, and new types of morphology; improving the constraint grammar to better produce forms which agree properly; and adding transfer rules for several new input configurations including associatives and adjectives.<br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for demonstrative pronouns/determiners (proximal, distal, and anaphoric)<br /> **Example word: &quot;rĩrĩa&quot; (Class 5 singular distal demonstrative)<br /> **Old output: &lt;code&gt;^rĩrĩa/*rĩrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^rĩrĩa/pro&lt;rel&gt;&lt;cl5&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dist&gt;&lt;cl5&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for possessive pronouns/determiners<br /> **Example word: &quot;gĩakwa&quot; (Possessive pronoun for Class 7 singular object with first-person singular owner)<br /> **Old output: &lt;code&gt;^gĩakwa/*gĩakwa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gĩakwa/pro&lt;prn&gt;&lt;poss&gt;&lt;cl7&gt;&lt;sg&gt;&lt;sp1&gt;&lt;ssg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for relative pronouns<br /> **Example word: ũrĩa (Class 1 singular relative pronoun, as well as other noun classes and other types of pronouns)<br /> **Old output: &lt;code&gt;^ũrĩa/*ũrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^ũrĩa/pro&lt;prn&gt;&lt;rel&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl14&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl14&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for personal pronouns<br /> **Example word: yo (Class 3 plural personal pronoun)<br /> **Old output: &lt;code&gt;^yo/*yo$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yo/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;sg&gt;$&lt;/code&gt;<br /> <br /> *Support for proper names (proper name tag introduced, and several names added)<br /> **Example word: Kenya (Kenya)<br /> **Old output: &lt;code&gt;^Kenya/Kenya$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kenya/Kenya&lt;n&gt;&lt;prop&gt;$&lt;/code&gt;<br /> <br /> *Support for Class 16 and 17 nouns (which tend to have to do with place), and their agreement on adjectives<br /> **Example: kũndũ guothe (all places)<br /> **Old output: &lt;code&gt;^kũndũ/*kũndũ$ ^guothe/*guothe$&lt;/code&gt;<br /> **New output: &lt;code&gt;^kũndũ/ndũ&lt;n&gt;&lt;cl17&gt;&lt;sg&gt;&lt;indef&gt;$ ^guothe/othe&lt;adj&gt;&lt;cl17&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for unique reversive middle voice morphology<br /> **Example word: rũthiũrũrũkĩirie (surround, roughly)<br /> **Old output: &lt;code&gt;^makĩinũkae/*makĩinũka$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^makĩinũka/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for infinitives<br /> **Example word: gũthambĩra (to swim)<br /> **Old output: &lt;code&gt;^gũthambĩra/*gũthambĩra$ ^guothe/*guothe$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gũthambĩra/^gũthambĩra/thambĩr&lt;v&gt;&lt;inf&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *An attempt was made to implement object agreement morphology on verbs. Such an attempt would necessitate adding twoc rules for each noun class, and doing so made twoc take far too long to compile. This attempt was therefore abandoned, but if a workaround were to be found for twoc, the twoc rules and lexc code are commented out in their respective files.<br /> <br /> ====Additional Smoothing of Transducer====<br /> <br /> *Support for ĩ to y alternation in certain cases (implemented through twol)<br /> **Example word: yarĩ (copula form)<br /> **Old output: &lt;code&gt;^yarĩ/*yarĩ$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yarĩ/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p2&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p13&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl7&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p10&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for ũ to u alternation before o<br /> **Example word: kuona (to see)<br /> **Old output: &lt;code&gt;^kuona/*kuona$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^kuona/on&lt;v&gt;&lt;inf&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> ====Additional Constraint Grammar Rules====<br /> <br /> Note: several of the rules in this section required multiple rlx rules to implement, because they act on various different aspects of agreement, including number, noun class, and person.<br /> <br /> *If I am a verb, demonstrative, or associative inflected to agree with a certain number/noun class, select me if I am preceded by an adjective preceded by a noun of that number/noun class.<br /> **Example: maaĩ maingĩ me (many waters are)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;Maaĩ&gt;&quot;<br /> &quot;aĩ&quot; n cl5 pl<br /> &quot;&lt;maingĩ&gt;&quot;<br /> &quot;ingĩ&quot; adj cl5 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl9p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl14 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl11p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl15 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl12p6 pl SELECT:27<br /> &quot;&lt;me&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:33<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/noun class, select me if I am preceded by a demonstrative preceded by a noun of that number/person/noun class.<br /> **Example: andũ acio nĩmagakorwo (those people will find)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;andũ&gt;&quot;<br /> &quot;ndũ&quot; n cl1 pl<br /> &quot;&lt;acio&gt;&quot;<br /> &quot;pro&quot; prn dem ana cl1 pl<br /> &quot;&lt;nĩmagakorwo&gt;&quot;<br /> &quot;kor&quot; v pres pass p3 cl1 pl foc SELECT:40<br /> &quot;kor&quot; v remfut pass p3 cl1 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl12p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl12p6 pl foc SELECT:40<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/person/noun class, select me if I am preceded by a number preceded by an associative preceded by a noun of that noun class.<br /> **Example: nyamũ ya mũthia nĩ (the last animal is)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;nyamũ&gt;&quot;<br /> &quot;nyamũ&quot; n cl9p10 pl<br /> &quot;nyamũ&quot; n cl9p10 sg<br /> &quot;&lt;ya&gt;&quot;<br /> &quot;a&quot; assoc cl9p10 sg SELECT:27<br /> ; &quot;a&quot; assoc cl3 pl SELECT:27<br /> ; &quot;a&quot; assoc cl9p6 sg SELECT:27<br /> &quot;&lt;mũthia&gt;&quot;<br /> &quot;mũthia&quot; num<br /> &quot;&lt;nĩ&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl9p10 pl SELECT:46<br /> &quot;rĩ&quot; cop pres p3 cl9p10 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 sg SELECT:46<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a preposition followed by a verb, remove me.<br /> **Example: na ndĩarĩ na kĩndũ (and was of nothing)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> ; &quot;na&quot; pr REMOVE:68<br /> &quot;&lt;ndĩarĩ&gt;&quot;<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p6 sg<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p10 sg<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;kĩndũ&gt;&quot;<br /> &quot;ndũ&quot; n cl7 sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> Note: None of the examples here will work, because there remains a bug in generation in the Kikuyu transducer. It cannot generate most forms it can analyze. All the forms outputted in these examples would be generated correctly if this bug is fixed, without other changes to the Kikuyu transducer.<br /> <br /> The chunker is the only part of the structural transfer system which has anything beyond the default, so only the chunker and final output is given.<br /> *Implemented transfer of adjectives<br /> **Example: ezibele monene<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^rango&lt;n&gt;&lt;cl3&gt;&lt;sg&gt;$ ^nene&lt;adj&gt;&lt;cl3&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}&lt;/code&gt;<br /> **Final output: &lt;code&gt; #rango #nene &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũrango mũnene &lt;/code&gt;<br /> <br /> *Implemented transfer of associative<br /> **Example: bato ya mokili (the people of the world)<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;pl&gt;$ ^a&lt;assoc&gt;&lt;cl1&gt;&lt;pl&gt;$ ^thĩ&lt;n&gt;&lt;cl9p10&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ a thĩ &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; andũ a thĩ &lt;/code&gt;<br /> <br /> *Implemented transfer of demonstratives<br /> **Example: moto óyo (this person)<br /> **Output of chunker: &lt;code&gt;^nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;sg&gt;$ ^pro&lt;prn&gt;&lt;dem&gt;&lt;prox&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ ũyũ&lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũndũ ũyũ &lt;/code&gt;<br /> <br /> ===kik → lin Implementation Details===<br /> This direction of the RBMT system was improved by adding items to the bilingual dictionary, adding additional morphology to the transducer, writing more constraint grammar rules, and writing additional transfer rules. Specific implementation details with examples of improved output are detailed in the sections that follow.<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for proper nouns<br /> **Example sentence: &quot;Nzambe&quot; (God)<br /> **Old output: &lt;code&gt;^Nzambe/*Nzambe$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nzambe/Nzambe&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for imperative mood<br /> **Example sentence: &quot;Kíma!&quot; (Run!)<br /> **Old output: &lt;code&gt;^Kíma/*Kíma$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kíma/kokíma&lt;v&gt;&lt;imp&gt;$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for intensifying suffixes (-a becomes -áká to denote certainty (infinitive, future) or insistence (imperative))<br /> **Example sentence 1: &quot;Azalákí kopésáká&quot; (He was really distributing...)<br /> **Old output: &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/*kopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/kopésa&lt;v&gt;&lt;inf&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 2: &quot;Akopésáká&quot; (He will really give...)<br /> **Old output: &lt;code&gt;^Akopésáká/*Akopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Akopésáká/kopésa&lt;v&gt;&lt;intns&gt;&lt;fut&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 3: &quot;Pésáká&quot; (Now do give...)<br /> **Old output: &lt;code&gt;^Pésáká/*Pésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Pésáká/kopésa&lt;v&gt;&lt;imp&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for more than 1 radical extension<br /> **Example sentence: &quot;Nabángisamákí na Pierre&quot; (I was scared by Pierre, or more literally, I was made, by Pierre, to fear something/someone)<br /> **Old output: &lt;code&gt;^Nabángisamákí/*Nabángisamákí$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nabángisamákí/kobánga&lt;v&gt;&lt;caus&gt;&lt;pass&gt;&lt;urp&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for numerals<br /> **Example sentence: &quot;zómi&quot; (ten)<br /> **Old output: &lt;code&gt;^zómi/*zómi$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^zómi/zómi&lt;num&gt;&lt;ord&gt;/zómi&lt;num&gt;&lt;card&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for quantifiers<br /> **Example sentence: &quot;mbelí nyónso&quot; (each knife)<br /> **Old output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/*nyónso$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/nyónso&lt;det&gt;&lt;qnt&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> ====Additional Constraint Grammar Rules====<br /> *If I am an adjective and a quantifier but I do not agree in number with the preceding noun, I must be a quantifier<br /> **Example sentence: &quot;bilóko moké&quot; (A few things, &#039;&#039;not&#039;&#039; small things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> &quot;moké&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> ; &quot;moké&quot; adj sg REMOVE:21<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a singular adjective, I cannot be plural<br /> **Example sentence: &quot;nzelá molaí&quot; (a long road)<br /> **Old output (below) <br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> ; &quot;nzelá&quot; n cl9 pl REMOVE:25<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a plural adjective, I cannot be singular<br /> **Example sentence: &quot;nzelá milaí&quot; (long roads)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 pl<br /> ; &quot;nzelá&quot; n cl9 sg REMOVE:28<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be an ordinal and a cardinal and am modifying a plural noun, I am not an ordinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;bilóko míbalé&quot; (two things, &#039;&#039;not&#039;&#039; the second thing)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num card<br /> ; &quot;míbalé&quot; num ord REMOVE:31<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be and ordinal and a cardinal and am modifying a a singular noun, I am not a cardinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;elóko mibalé&quot; (the second thing, &#039;&#039;not&#039;&#039; two things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> ; &quot;míbalé&quot; num card REMOVE:34<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be a demonstrative nominal and a relative pronoun and am proceeded by a noun, I cannot be a demonstrative nominal.<br /> **Example sentence: &quot;Mobali óyo amonaki ngáí&quot; (The man who saw me...)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> &quot;óyo&quot; prn dem sg<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> ; &quot;óyo&quot; prn dem sg REMOVE:37<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> *Enhance existing basic verb transfer rule to work with newly implemented imperative<br /> **Example sentence: &quot;ina&quot; (sing)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#koyemba&lt;/code&gt;<br /> **New output<br /> ***Tagger:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;imp&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;yemba&lt;/code&gt;<br /> <br /> *Map Kikuyu associative to Lingala connective by removing extra agreement tags found in Kikuyu<br /> **Example sentence: &quot;wa&quot;<br /> **Old output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#ya&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;ya&lt;/code&gt;<br /> <br /> *Basic adjective transfer: keep number markings if a variable adjective in Lingala, discard otherwise<br /> **Example sentence 1: anene (big)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#monéne&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;minéne&lt;/code&gt;<br /> **Example sentence 2: mata (sour)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#bololo&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;bololo&lt;/code&gt;<br /> <br /> ===lin Transducer Evaluation===<br /> *Precision (annotated corpus): 80.53%<br /> *Recall (annotated corpus): 55.64%<br /> *Coverage (large corpus): 62.10%<br /> *Number of words (large corpus): 275,546 (actual), 304,425 (tokenized)<br /> *Number of stems in transducer: 273<br /> <br /> ===kik Transducer Evaluation===<br /> *Precision (annotated corpus):<br /> *Recall (annotated corpus): <br /> *Coverage (large corpus): 58.39%<br /> *Number of words (large corpus): 107564<br /> *Number of stems in transducer: about 267 (nouns, verbs, adjectives, and adverbs)<br /> <br /> ===lin → kik MT Evaluation===<br /> Note that the bug in the Kikuyu generator significantly alters these numbers. They should improve significantly if the bug is resolved.<br /> *WER (longer corpus): 96.49%<br /> *PER (longer corpus): 92.98%<br /> *Percentage unknown words: 13.91%<br /> *Percentage known words: 86.09%<br /> *Trimmed coverage over longer corpus [lin-kik.automorf.bin applied to lin.longer.txt]: 89.29%<br /> *Trimmed coverage over large corpus [lin-kik.automorf.bin applied to lin.corpus.large.txt]: 48.90%<br /> *Number of stems in longer corpus: 117 (actual) 140 (tokenized, perhaps including punctuation)<br /> *Number of stems in large corpus: 275,546 (actual), 304,425 (tokenized)<br /> <br /> ===kik → lin MT Evaluation===<br /> *WER (longer corpus): 86.32%<br /> *PER (longer corpus): 75.21%<br /> *Percentage unknown words: 39.50% [Actually probably a bit higher since some words were marked with an &quot;@&quot; and not translated while the script only looks for words marked with a &quot;*&quot;]<br /> *Percentage known words: 60.60%<br /> *Trimmed coverage over longer corpus [kik-lin.automorf.bin applied to kik.corpus.larger.txt]: 71.43%<br /> *Trimmed coverage over large corpus [kik-lin.automorf.bin applied to kik.longer.txt]: 46.40%<br /> *Number of stems in longer corpus: 115<br /> *Number of stems in large corpus: 107564<br /> <br /> <br /> [[Category:Sp17_TranslationPairs]]<br /> [[Category:Lingala]]<br /> [[Category:Kikuyu]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Lingala_and_Kikuyu&diff=4586 2017-04-18T01:06:55Z

<p>Jmundo1: </p> <hr /> <div>Resources for machine translation between [[Lingala]] and [[Kikuyu]]. Code is stored in a [https://github.swarthmore.edu/apowell1/ling073-kik-lin GitHub Repository].<br /> ==Developed Resources==<br /> ===Parallel Corpus===<br /> <br /> A [https://github.swarthmore.edu/apowell1/ling073-lin-kik-corpus parallel corpus] created using two translations of Genesis is available on Github.<br /> <br /> ===Contrastive Grammar===<br /> Contrastive Grammar is detailed on the wiki page [[Lingala and Kikuyu/Contrastive Grammar]].<br /> <br /> ===Lexical Selection===<br /> Lexical selection is detailed on the wiki page [[Lingala and Kikuyu/Lexical Selection]].<br /> <br /> ===Structural Transfer===<br /> Structural transfer is detailed on the wiki page [[Lingala and Kikuyu/Structural Transfer]]<br /> <br /> ==Installing Unidecode For MT Evaluation==<br /> Most available Lingala text does not mark tones; however, the Lingala transducer has been configured using spellrelax to provide analyses and surface generation with tone markings included as accents on vowels. As a result, generated translations will include tone markings not present in the reference translation and artificially inflate evaluation metrics such as WER and PER. Fortunately, we can take advantage of the fact that accented characters are not present in ASCII to replace accented characters with their unaccented version on the command line.<br /> <br /> [https://pypi.python.org/pypi/Unidecode Unidecode] is a Python package that implements a function to convert Unicode strings into their closest ASCII approximation. For simple accented characters, this is precisely the mapping we need to remove the accents for evaluation against a reference translation. Unidecode can also be used on the command line with our machine translation system as follows: &lt;code&gt;echo /path/to/kik-corpus | apertium -d . kik-lin | unidecode&lt;/code&gt;. To install unidecode, it is recommended to use &lt;code&gt;pip&lt;/code&gt;, as experience suggests that &lt;code&gt;apt-get&lt;/code&gt; will not run the necessary installation scripts to make unidecode accessible in Bash.<br /> <br /> On the Ling 073 virtual machines, Pip can be installed using &lt;code&gt;sudo apt-get install python-pip&lt;/code&gt;. After installing pip, unidecode can be installed using &lt;code&gt; pip install unidecode&lt;/code&gt;. <br /> <br /> ==Lexical Selection==<br /> This section is part of the [[Lexical transfer]] assignment. In addition to adding many one-to-one mappings to our .dix file, we have included a couple of one-to-many mappings for each direction of translation. See [[Lingala and Kikuyu/Lexical Selection]] for further details.<br /> <br /> From Lingala to Kikuyu, we consider four one-to-many mappings:<br /> *The Lingala verb kokíma can be translated in English as both &quot;to run&quot; and &quot;to run away.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, hanyũk and ũr.<br /> *The Lingala verb kofánda can be translated in English as both &quot;to sit&quot; and &quot;to live&quot; (as in reside). Kikuyu has two distinct verbs corresponding to these English translations, respectively, ikar and tũũr.<br /> *The Lingala verb kobánza can be translated in English as both &quot;to think&quot; and &quot;to remember.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, ĩcir and ririkan.<br /> *The Lingala noun mabele can be translated in English as both &quot;sand&quot; and &quot;milk.&quot; Kikuyu has two distinct nouns corresponding to these English translations, respectively, thanga and ria.<br /> <br /> From Kikuyu to Lingala, we consider three one-to-many-mappings:<br /> *The Kikuyu verb igu can be translated in English as both &quot;to obey&quot; and &quot;to hear.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, kotosa and koyoka.<br /> *The Kikuyu verb rut can be translated in English as both &quot;to learn&quot; and &quot;to remove.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, koyekola and kokukula.<br /> *the Kikuyu noun ria can be translated in English as both &quot;milk&quot; and &quot;pond.&quot; Lingala has two distinct nouns corresponding to these English translations, respectively, mabele and etimá.<br /> <br /> ==Initial Evaluation==<br /> ===lin → kik evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 87.50%<br /> **PER: 87.50%<br /> **Number of words: 9<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 99.06%<br /> **PER: 94.34%<br /> **Number of words: 107<br /> **Number of unknown words: 71<br /> **Proportion of stems translated correctly:33.64%<br /> <br /> ===kik → lin evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 80%<br /> **PER: 80%<br /> **Number of words: 7 (This should really be 8, but the command we are supposed to use to count this outputs 7. The proportion remains unchanged in our case, however.)<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 98.13%<br /> **PER: 90.65%<br /> **Number of words: 106<br /> **Number of unknown words: 79 (Note: Using the commands on the wiki gives a slightly different count that using the apertium-eval-translator tool. We use the count given by the command, not the tool)<br /> **Proportion of stems translated correctly: 25.47%<br /> <br /> ==Final Evaluation==<br /> ===lin → kik Implementation Details===<br /> <br /> This direction of the pair was improved by improving the transducer to handle pronouns, proper names, and new types of morphology; improving the constraint grammar to better produce forms which agree properly; and adding transfer rules for several new input configurations including associatives and adjectives.<br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for demonstrative pronouns/determiners (proximal, distal, and anaphoric)<br /> **Example word: &quot;rĩrĩa&quot; (Class 5 singular distal demonstrative)<br /> **Old output: &lt;code&gt;^rĩrĩa/*rĩrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^rĩrĩa/pro&lt;rel&gt;&lt;cl5&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dist&gt;&lt;cl5&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for possessive pronouns/determiners<br /> **Example word: &quot;gĩakwa&quot; (Possessive pronoun for Class 7 singular object with first-person singular owner)<br /> **Old output: &lt;code&gt;^gĩakwa/*gĩakwa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gĩakwa/pro&lt;prn&gt;&lt;poss&gt;&lt;cl7&gt;&lt;sg&gt;&lt;sp1&gt;&lt;ssg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for relative pronouns<br /> **Example word: ũrĩa (Class 1 singular relative pronoun, as well as other noun classes and other types of pronouns)<br /> **Old output: &lt;code&gt;^ũrĩa/*ũrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^ũrĩa/pro&lt;prn&gt;&lt;rel&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl14&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl14&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for personal pronouns<br /> **Example word: yo (Class 3 plural personal pronoun)<br /> **Old output: &lt;code&gt;^yo/*yo$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yo/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;sg&gt;$&lt;/code&gt;<br /> <br /> *Support for proper names (proper name tag introduced, and several names added)<br /> **Example word: Kenya (Kenya)<br /> **Old output: &lt;code&gt;^Kenya/Kenya$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kenya/Kenya&lt;n&gt;&lt;prop&gt;$&lt;/code&gt;<br /> <br /> *Support for Class 16 and 17 nouns (which tend to have to do with place), and their agreement on adjectives<br /> **Example: kũndũ guothe (all places)<br /> **Old output: &lt;code&gt;^kũndũ/*kũndũ$ ^guothe/*guothe$&lt;/code&gt;<br /> **New output: &lt;code&gt;^kũndũ/ndũ&lt;n&gt;&lt;cl17&gt;&lt;sg&gt;&lt;indef&gt;$ ^guothe/othe&lt;adj&gt;&lt;cl17&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for unique reversive middle voice morphology<br /> **Example word: rũthiũrũrũkĩirie (surround, roughly)<br /> **Old output: &lt;code&gt;^makĩinũkae/*makĩinũka$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^makĩinũka/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for infinitives<br /> **Example word: gũthambĩra (to swim)<br /> **Old output: &lt;code&gt;^gũthambĩra/*gũthambĩra$ ^guothe/*guothe$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gũthambĩra/^gũthambĩra/thambĩr&lt;v&gt;&lt;inf&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *An attempt was made to implement object agreement morphology on verbs. Such an attempt would necessitate adding twoc rules for each noun class, and doing so made twoc take far too long to compile. This attempt was therefore abandoned, but if a workaround were to be found for twoc, the twoc rules and lexc code are commented out in their respective files.<br /> <br /> ====Additional Smoothing of Transducer====<br /> <br /> *Support for ĩ to y alternation in certain cases (implemented through twol)<br /> **Example word: yarĩ (copula form)<br /> **Old output: &lt;code&gt;^yarĩ/*yarĩ$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yarĩ/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p2&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p13&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl7&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p10&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for ũ to u alternation before o<br /> **Example word: kuona (to see)<br /> **Old output: &lt;code&gt;^kuona/*kuona$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^kuona/on&lt;v&gt;&lt;inf&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> ====Additional Constraint Grammar Rules====<br /> <br /> Note: several of the rules in this section required multiple rlx rules to implement, because they act on various different aspects of agreement, including number, noun class, and person.<br /> <br /> *If I am a verb, demonstrative, or associative inflected to agree with a certain number/noun class, select me if I am preceded by an adjective preceded by a noun of that number/noun class.<br /> **Example: maaĩ maingĩ me (many waters are)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;Maaĩ&gt;&quot;<br /> &quot;aĩ&quot; n cl5 pl<br /> &quot;&lt;maingĩ&gt;&quot;<br /> &quot;ingĩ&quot; adj cl5 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl9p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl14 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl11p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl15 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl12p6 pl SELECT:27<br /> &quot;&lt;me&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:33<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/noun class, select me if I am preceded by a demonstrative preceded by a noun of that number/person/noun class.<br /> **Example: andũ acio nĩmagakorwo (those people will find)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;andũ&gt;&quot;<br /> &quot;ndũ&quot; n cl1 pl<br /> &quot;&lt;acio&gt;&quot;<br /> &quot;pro&quot; prn dem ana cl1 pl<br /> &quot;&lt;nĩmagakorwo&gt;&quot;<br /> &quot;kor&quot; v pres pass p3 cl1 pl foc SELECT:40<br /> &quot;kor&quot; v remfut pass p3 cl1 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl12p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl12p6 pl foc SELECT:40<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/person/noun class, select me if I am preceded by a number preceded by an associative preceded by a noun of that noun class.<br /> **Example: nyamũ ya mũthia nĩ (the last animal is)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;nyamũ&gt;&quot;<br /> &quot;nyamũ&quot; n cl9p10 pl<br /> &quot;nyamũ&quot; n cl9p10 sg<br /> &quot;&lt;ya&gt;&quot;<br /> &quot;a&quot; assoc cl9p10 sg SELECT:27<br /> ; &quot;a&quot; assoc cl3 pl SELECT:27<br /> ; &quot;a&quot; assoc cl9p6 sg SELECT:27<br /> &quot;&lt;mũthia&gt;&quot;<br /> &quot;mũthia&quot; num<br /> &quot;&lt;nĩ&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl9p10 pl SELECT:46<br /> &quot;rĩ&quot; cop pres p3 cl9p10 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 sg SELECT:46<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a preposition followed by a verb, remove me.<br /> **Example: na ndĩarĩ na kĩndũ (and was of nothing)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> ; &quot;na&quot; pr REMOVE:68<br /> &quot;&lt;ndĩarĩ&gt;&quot;<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p6 sg<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p10 sg<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;kĩndũ&gt;&quot;<br /> &quot;ndũ&quot; n cl7 sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> Note: None of the examples here will work, because there remains a bug in generation in the Kikuyu transducer. It cannot generate most forms it can analyze. All the forms outputted in these examples would be generated correctly if this bug is fixed, without other changes to the Kikuyu transducer.<br /> <br /> The chunker is the only part of the structural transfer system which has anything beyond the default, so only the chunker and final output is given.<br /> *Implemented transfer of adjectives<br /> **Example: ezibele monene<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^rango&lt;n&gt;&lt;cl3&gt;&lt;sg&gt;$ ^nene&lt;adj&gt;&lt;cl3&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}&lt;/code&gt;<br /> **Final output: &lt;code&gt; #rango #nene &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũrango mũnene &lt;/code&gt;<br /> <br /> *Implemented transfer of associative<br /> **Example: bato ya mokili (the people of the world)<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;pl&gt;$ ^a&lt;assoc&gt;&lt;cl1&gt;&lt;pl&gt;$ ^thĩ&lt;n&gt;&lt;cl9p10&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ a thĩ &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; andũ a thĩ &lt;/code&gt;<br /> <br /> *Implemented transfer of demonstratives<br /> **Example: moto óyo (this person)<br /> **Output of chunker: &lt;code&gt;^nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;sg&gt;$ ^pro&lt;prn&gt;&lt;dem&gt;&lt;prox&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ ũyũ&lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũndũ ũyũ &lt;/code&gt;<br /> <br /> ===kik → lin Implementation Details===<br /> This direction of the RBMT system was improved by adding items to the bilingual dictionary, adding additional morphology to the transducer, writing more constraint grammar rules, and writing additional transfer rules. Specific implementation details with examples of improved output are detailed in the sections that follow.<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for proper nouns<br /> **Example sentence: &quot;Nzambe&quot; (God)<br /> **Old output: &lt;code&gt;^Nzambe/*Nzambe$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nzambe/Nzambe&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for imperative mood<br /> **Example sentence: &quot;Kíma!&quot; (Run!)<br /> **Old output: &lt;code&gt;^Kíma/*Kíma$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kíma/kokíma&lt;v&gt;&lt;imp&gt;$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for intensifying suffixes (-a becomes -áká to denote certainty (infinitive, future) or insistence (imperative))<br /> **Example sentence 1: &quot;Azalákí kopésáká&quot; (He was really distributing...)<br /> **Old output: &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/*kopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/kopésa&lt;v&gt;&lt;inf&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 2: &quot;Akopésáká&quot; (He will really give...)<br /> **Old output: &lt;code&gt;^Akopésáká/*Akopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Akopésáká/kopésa&lt;v&gt;&lt;intns&gt;&lt;fut&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 3: &quot;Pésáká&quot; (Now do give...)<br /> **Old output: &lt;code&gt;^Pésáká/*Pésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Pésáká/kopésa&lt;v&gt;&lt;imp&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for more than 1 radical extension<br /> **Example sentence: &quot;Nabángisamákí na Pierre&quot; (I was scared by Pierre, or more literally, I was made, by Pierre, to fear something/someone)<br /> **Old output: &lt;code&gt;^Nabángisamákí/*Nabángisamákí$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nabángisamákí/kobánga&lt;v&gt;&lt;caus&gt;&lt;pass&gt;&lt;urp&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for numerals<br /> **Example sentence: &quot;zómi&quot; (ten)<br /> **Old output: &lt;code&gt;^zómi/*zómi$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^zómi/zómi&lt;num&gt;&lt;ord&gt;/zómi&lt;num&gt;&lt;card&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for quantifiers<br /> **Example sentence: &quot;mbelí nyónso&quot; (each knife)<br /> **Old output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/*nyónso$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/nyónso&lt;det&gt;&lt;qnt&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> ====Additional Constraint Grammar Rules====<br /> *If I am an adjective and a quantifier but I do not agree in number with the preceding noun, I must be a quantifier<br /> **Example sentence: &quot;bilóko moké&quot; (A few things, &#039;&#039;not&#039;&#039; small things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> &quot;moké&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> ; &quot;moké&quot; adj sg REMOVE:21<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a singular adjective, I cannot be plural<br /> **Example sentence: &quot;nzelá molaí&quot; (a long road)<br /> **Old output (below) <br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> ; &quot;nzelá&quot; n cl9 pl REMOVE:25<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a plural adjective, I cannot be singular<br /> **Example sentence: &quot;nzelá milaí&quot; (long roads)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 pl<br /> ; &quot;nzelá&quot; n cl9 sg REMOVE:28<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be an ordinal and a cardinal and am modifying a plural noun, I am not an ordinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;bilóko míbalé&quot; (two things, &#039;&#039;not&#039;&#039; the second thing)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num card<br /> ; &quot;míbalé&quot; num ord REMOVE:31<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be and ordinal and a cardinal and am modifying a a singular noun, I am not a cardinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;elóko mibalé&quot; (the second thing, &#039;&#039;not&#039;&#039; two things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> ; &quot;míbalé&quot; num card REMOVE:34<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be a demonstrative nominal and a relative pronoun and am proceeded by a noun, I cannot be a demonstrative nominal.<br /> **Example sentence: &quot;Mobali óyo amonaki ngáí&quot; (The man who saw me...)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> &quot;óyo&quot; prn dem sg<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> ; &quot;óyo&quot; prn dem sg REMOVE:37<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> *Enhance existing basic verb transfer rule to work with newly implemented imperative<br /> **Example sentence: &quot;ina&quot; (sing)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#koyemba&lt;/code&gt;<br /> **New output<br /> ***Tagger:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;imp&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;yemba&lt;/code&gt;<br /> <br /> *Map Kikuyu associative to Lingala connective by removing extra agreement tags found in Kikuyu<br /> **Example sentence: &quot;wa&quot;<br /> **Old output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#ya&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;ya&lt;/code&gt;<br /> <br /> *Basic adjective transfer: keep number markings if a variable adjective in Lingala, discard otherwise<br /> **Example sentence 1: anene (big)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#monéne&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;minéne&lt;/code&gt;<br /> **Example sentence 2: mata (sour)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#bololo&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;bololo&lt;/code&gt;<br /> <br /> ===lin Transducer Evaluation===<br /> *Precision (annotated corpus): 80.53%<br /> *Recall (annotated corpus): 55.64%<br /> *Coverage (large corpus): 62.10%<br /> *Number of words (large corpus): 275,546 (actual), 304,425 (tokenized)<br /> *Number of stems in transducer: 273<br /> <br /> ===kik Transducer Evaluation===<br /> *Precision (annotated corpus):<br /> *Recall (annotated corpus): <br /> *Coverage (large corpus):<br /> *Number of words (large corpus):<br /> *Number of stems in transducer:<br /> <br /> ===lin → kik MT Evaluation===<br /> *WER (longer corpus):<br /> *PER (longer corpus):<br /> *Percentage unknown words:<br /> *Percentage known words:<br /> *Trimmed coverage over longer corpus [lin-kik.automorf.bin applied to lin.longer.txt]: 89.29%<br /> *Trimmed coverage over large corpus [lin-kik.automorf.bin applied to lin.corpus.large.txt]: 48.90%<br /> *Number of stems in longer corpus: 117 (actual) 140 (tokenized, perhaps including punctuation)<br /> *Number of stems in large corpus: 275,546 (actual), 304,425 (tokenized)<br /> <br /> ===kik → lin MT Evaluation===<br /> *WER (longer corpus): 86.32%<br /> *PER (longer corpus): 75.21%<br /> *Percentage unknown words: 39.50% [Actually probably a bit higher since some words were marked with an &quot;@&quot; and not translated while the script only looks for words marked with a &quot;*&quot;]<br /> *Percentage known words: 60.60%<br /> *Trimmed coverage over longer corpus [kik-lin.automorf.bin applied to kik.corpus.larger.txt]:<br /> *Trimmed coverage over large corpus [kik-lin.automorf.bin applied to kik.longer.txt]:<br /> *Number of stems in longer corpus:<br /> *Number of stems in large corpus:<br /> <br /> <br /> [[Category:Sp17_TranslationPairs]]<br /> [[Category:Lingala]]<br /> [[Category:Kikuyu]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Lingala_and_Kikuyu&diff=4585 2017-04-18T01:04:09Z

<p>Jmundo1: </p> <hr /> <div>Resources for machine translation between [[Lingala]] and [[Kikuyu]]. Code is stored in a [https://github.swarthmore.edu/apowell1/ling073-kik-lin GitHub Repository].<br /> ==Developed Resources==<br /> ===Parallel Corpus===<br /> <br /> A [https://github.swarthmore.edu/apowell1/ling073-lin-kik-corpus parallel corpus] created using two translations of Genesis is available on Github.<br /> <br /> ===Contrastive Grammar===<br /> Contrastive Grammar is detailed on the wiki page [[Lingala and Kikuyu/Contrastive Grammar]].<br /> <br /> ===Lexical Selection===<br /> Lexical selection is detailed on the wiki page [[Lingala and Kikuyu/Lexical Selection]].<br /> <br /> ===Structural Transfer===<br /> Structural transfer is detailed on the wiki page [[Lingala and Kikuyu/Structural Transfer]]<br /> <br /> ==Installing Unidecode For MT Evaluation==<br /> Most available Lingala text does not mark tones; however, the Lingala transducer has been configured using spellrelax to provide analyses and surface generation with tone markings included as accents on vowels. As a result, generated translations will include tone markings not present in the reference translation and artificially inflate evaluation metrics such as WER and PER. Fortunately, we can take advantage of the fact that accented characters are not present in ASCII to replace accented characters with their unaccented version on the command line.<br /> <br /> [https://pypi.python.org/pypi/Unidecode Unidecode] is a Python package that implements a function to convert Unicode strings into their closest ASCII approximation. For simple accented characters, this is precisely the mapping we need to remove the accents for evaluation against a reference translation. Unidecode can also be used on the command line with our machine translation system as follows: &lt;code&gt;echo /path/to/kik-corpus | apertium -d . kik-lin | unidecode&lt;/code&gt;. To install unidecode, it is recommended to use &lt;code&gt;pip&lt;/code&gt;, as experience suggests that &lt;code&gt;apt-get&lt;/code&gt; will not run the necessary installation scripts to make unidecode accessible in Bash.<br /> <br /> On the Ling 073 virtual machines, Pip can be installed using &lt;code&gt;sudo apt-get install python-pip&lt;/code&gt;. After installing pip, unidecode can be installed using &lt;code&gt; pip install unidecode&lt;/code&gt;. <br /> <br /> ==Lexical Selection==<br /> This section is part of the [[Lexical transfer]] assignment. In addition to adding many one-to-one mappings to our .dix file, we have included a couple of one-to-many mappings for each direction of translation. See [[Lingala and Kikuyu/Lexical Selection]] for further details.<br /> <br /> From Lingala to Kikuyu, we consider four one-to-many mappings:<br /> *The Lingala verb kokíma can be translated in English as both &quot;to run&quot; and &quot;to run away.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, hanyũk and ũr.<br /> *The Lingala verb kofánda can be translated in English as both &quot;to sit&quot; and &quot;to live&quot; (as in reside). Kikuyu has two distinct verbs corresponding to these English translations, respectively, ikar and tũũr.<br /> *The Lingala verb kobánza can be translated in English as both &quot;to think&quot; and &quot;to remember.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, ĩcir and ririkan.<br /> *The Lingala noun mabele can be translated in English as both &quot;sand&quot; and &quot;milk.&quot; Kikuyu has two distinct nouns corresponding to these English translations, respectively, thanga and ria.<br /> <br /> From Kikuyu to Lingala, we consider three one-to-many-mappings:<br /> *The Kikuyu verb igu can be translated in English as both &quot;to obey&quot; and &quot;to hear.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, kotosa and koyoka.<br /> *The Kikuyu verb rut can be translated in English as both &quot;to learn&quot; and &quot;to remove.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, koyekola and kokukula.<br /> *the Kikuyu noun ria can be translated in English as both &quot;milk&quot; and &quot;pond.&quot; Lingala has two distinct nouns corresponding to these English translations, respectively, mabele and etimá.<br /> <br /> ==Initial Evaluation==<br /> ===lin → kik evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 87.50%<br /> **PER: 87.50%<br /> **Number of words: 9<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 99.06%<br /> **PER: 94.34%<br /> **Number of words: 107<br /> **Number of unknown words: 71<br /> **Proportion of stems translated correctly:33.64%<br /> <br /> ===kik → lin evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 80%<br /> **PER: 80%<br /> **Number of words: 7 (This should really be 8, but the command we are supposed to use to count this outputs 7. The proportion remains unchanged in our case, however.)<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 98.13%<br /> **PER: 90.65%<br /> **Number of words: 106<br /> **Number of unknown words: 79 (Note: Using the commands on the wiki gives a slightly different count that using the apertium-eval-translator tool. We use the count given by the command, not the tool)<br /> **Proportion of stems translated correctly: 25.47%<br /> <br /> ==Final Evaluation==<br /> ===lin → kik Implementation Details===<br /> <br /> This direction of the pair was improved by improving the transducer to handle pronouns, proper names, and new types of morphology; improving the constraint grammar to better produce forms which agree properly; and adding transfer rules for several new input configurations including associatives and adjectives.<br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for demonstrative pronouns/determiners (proximal, distal, and anaphoric)<br /> **Example word: &quot;rĩrĩa&quot; (Class 5 singular distal demonstrative)<br /> **Old output: &lt;code&gt;^rĩrĩa/*rĩrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^rĩrĩa/pro&lt;rel&gt;&lt;cl5&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dist&gt;&lt;cl5&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for possessive pronouns/determiners<br /> **Example word: &quot;gĩakwa&quot; (Possessive pronoun for Class 7 singular object with first-person singular owner)<br /> **Old output: &lt;code&gt;^gĩakwa/*gĩakwa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gĩakwa/pro&lt;prn&gt;&lt;poss&gt;&lt;cl7&gt;&lt;sg&gt;&lt;sp1&gt;&lt;ssg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for relative pronouns<br /> **Example word: ũrĩa (Class 1 singular relative pronoun, as well as other noun classes and other types of pronouns)<br /> **Old output: &lt;code&gt;^ũrĩa/*ũrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^ũrĩa/pro&lt;prn&gt;&lt;rel&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl14&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl14&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for personal pronouns<br /> **Example word: yo (Class 3 plural personal pronoun)<br /> **Old output: &lt;code&gt;^yo/*yo$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yo/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;sg&gt;$&lt;/code&gt;<br /> <br /> *Support for proper names (proper name tag introduced, and several names added)<br /> **Example word: Kenya (Kenya)<br /> **Old output: &lt;code&gt;^Kenya/Kenya$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kenya/Kenya&lt;n&gt;&lt;prop&gt;$&lt;/code&gt;<br /> <br /> *Support for Class 16 and 17 nouns (which tend to have to do with place), and their agreement on adjectives<br /> **Example: kũndũ guothe (all places)<br /> **Old output: &lt;code&gt;^kũndũ/*kũndũ$ ^guothe/*guothe$&lt;/code&gt;<br /> **New output: &lt;code&gt;^kũndũ/ndũ&lt;n&gt;&lt;cl17&gt;&lt;sg&gt;&lt;indef&gt;$ ^guothe/othe&lt;adj&gt;&lt;cl17&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for unique reversive middle voice morphology<br /> **Example word: rũthiũrũrũkĩirie (surround, roughly)<br /> **Old output: &lt;code&gt;^makĩinũkae/*makĩinũka$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^makĩinũka/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for infinitives<br /> **Example word: gũthambĩra (to swim)<br /> **Old output: &lt;code&gt;^gũthambĩra/*gũthambĩra$ ^guothe/*guothe$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gũthambĩra/^gũthambĩra/thambĩr&lt;v&gt;&lt;inf&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *An attempt was made to implement object agreement morphology on verbs. Such an attempt would necessitate adding twoc rules for each noun class, and doing so made twoc take far too long to compile. This attempt was therefore abandoned, but if a workaround were to be found for twoc, the twoc rules and lexc code are commented out in their respective files.<br /> <br /> ====Additional Smoothing of Transducer====<br /> <br /> *Support for ĩ to y alternation in certain cases (implemented through twol)<br /> **Example word: yarĩ (copula form)<br /> **Old output: &lt;code&gt;^yarĩ/*yarĩ$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yarĩ/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p2&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p13&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl7&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p10&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for ũ to u alternation before o<br /> **Example word: kuona (to see)<br /> **Old output: &lt;code&gt;^kuona/*kuona$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^kuona/on&lt;v&gt;&lt;inf&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> ====Additional Constraint Grammar Rules====<br /> <br /> *If I am a verb, demonstrative, or associative inflected to agree with a certain number/noun class, select me if I am preceded by an adjective preceded by a noun of that number/noun class.<br /> **Example: maaĩ maingĩ me (many waters are)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;Maaĩ&gt;&quot;<br /> &quot;aĩ&quot; n cl5 pl<br /> &quot;&lt;maingĩ&gt;&quot;<br /> &quot;ingĩ&quot; adj cl5 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl9p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl14 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl11p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl15 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl12p6 pl SELECT:27<br /> &quot;&lt;me&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:33<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/noun class, select me if I am preceded by a demonstrative preceded by a noun of that number/person/noun class.<br /> **Example: andũ acio nĩmagakorwo (those people will find)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;andũ&gt;&quot;<br /> &quot;ndũ&quot; n cl1 pl<br /> &quot;&lt;acio&gt;&quot;<br /> &quot;pro&quot; prn dem ana cl1 pl<br /> &quot;&lt;nĩmagakorwo&gt;&quot;<br /> &quot;kor&quot; v pres pass p3 cl1 pl foc SELECT:40<br /> &quot;kor&quot; v remfut pass p3 cl1 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl12p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl12p6 pl foc SELECT:40<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/person/noun class, select me if I am preceded by a number preceded by an associative preceded by a noun of that noun class.<br /> **Example: nyamũ ya mũthia nĩ (the last animal is)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;nyamũ&gt;&quot;<br /> &quot;nyamũ&quot; n cl9p10 pl<br /> &quot;nyamũ&quot; n cl9p10 sg<br /> &quot;&lt;ya&gt;&quot;<br /> &quot;a&quot; assoc cl9p10 sg SELECT:27<br /> ; &quot;a&quot; assoc cl3 pl SELECT:27<br /> ; &quot;a&quot; assoc cl9p6 sg SELECT:27<br /> &quot;&lt;mũthia&gt;&quot;<br /> &quot;mũthia&quot; num<br /> &quot;&lt;nĩ&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl9p10 pl SELECT:46<br /> &quot;rĩ&quot; cop pres p3 cl9p10 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 sg SELECT:46<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a preposition followed by a verb, remove me.<br /> **Example: na ndĩarĩ na kĩndũ (and was of nothing)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> ; &quot;na&quot; pr REMOVE:68<br /> &quot;&lt;ndĩarĩ&gt;&quot;<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p6 sg<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p10 sg<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;kĩndũ&gt;&quot;<br /> &quot;ndũ&quot; n cl7 sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> Note: None of the examples here will work, because there remains a bug in generation in the Kikuyu transducer. It cannot generate most forms it can analyze. All the forms outputted in these examples would be generated correctly if this bug is fixed, without other changes to the Kikuyu transducer.<br /> <br /> The chunker is the only part of the structural transfer system which has anything beyond the default, so only the chunker and final output is given.<br /> *Implemented transfer of adjectives<br /> **Example: ezibele monene<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^rango&lt;n&gt;&lt;cl3&gt;&lt;sg&gt;$ ^nene&lt;adj&gt;&lt;cl3&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}&lt;/code&gt;<br /> **Final output: &lt;code&gt; #rango #nene &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũrango mũnene &lt;/code&gt;<br /> <br /> *Implemented transfer of associative<br /> **Example: bato ya mokili (the people of the world)<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;pl&gt;$ ^a&lt;assoc&gt;&lt;cl1&gt;&lt;pl&gt;$ ^thĩ&lt;n&gt;&lt;cl9p10&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ a thĩ &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; andũ a thĩ &lt;/code&gt;<br /> <br /> *Implemented transfer of demonstratives<br /> **Example: moto óyo (this person)<br /> **Output of chunker: &lt;code&gt;^nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;sg&gt;$ ^pro&lt;prn&gt;&lt;dem&gt;&lt;prox&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ ũyũ&lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũndũ ũyũ &lt;/code&gt;<br /> <br /> ===kik → lin Implementation Details===<br /> This direction of the RBMT system was improved by adding items to the bilingual dictionary, adding additional morphology to the transducer, writing more constraint grammar rules, and writing additional transfer rules. Specific implementation details with examples of improved output are detailed in the sections that follow.<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for proper nouns<br /> **Example sentence: &quot;Nzambe&quot; (God)<br /> **Old output: &lt;code&gt;^Nzambe/*Nzambe$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nzambe/Nzambe&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for imperative mood<br /> **Example sentence: &quot;Kíma!&quot; (Run!)<br /> **Old output: &lt;code&gt;^Kíma/*Kíma$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kíma/kokíma&lt;v&gt;&lt;imp&gt;$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for intensifying suffixes (-a becomes -áká to denote certainty (infinitive, future) or insistence (imperative))<br /> **Example sentence 1: &quot;Azalákí kopésáká&quot; (He was really distributing...)<br /> **Old output: &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/*kopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/kopésa&lt;v&gt;&lt;inf&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 2: &quot;Akopésáká&quot; (He will really give...)<br /> **Old output: &lt;code&gt;^Akopésáká/*Akopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Akopésáká/kopésa&lt;v&gt;&lt;intns&gt;&lt;fut&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 3: &quot;Pésáká&quot; (Now do give...)<br /> **Old output: &lt;code&gt;^Pésáká/*Pésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Pésáká/kopésa&lt;v&gt;&lt;imp&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for more than 1 radical extension<br /> **Example sentence: &quot;Nabángisamákí na Pierre&quot; (I was scared by Pierre, or more literally, I was made, by Pierre, to fear something/someone)<br /> **Old output: &lt;code&gt;^Nabángisamákí/*Nabángisamákí$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nabángisamákí/kobánga&lt;v&gt;&lt;caus&gt;&lt;pass&gt;&lt;urp&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for numerals<br /> **Example sentence: &quot;zómi&quot; (ten)<br /> **Old output: &lt;code&gt;^zómi/*zómi$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^zómi/zómi&lt;num&gt;&lt;ord&gt;/zómi&lt;num&gt;&lt;card&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for quantifiers<br /> **Example sentence: &quot;mbelí nyónso&quot; (each knife)<br /> **Old output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/*nyónso$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/nyónso&lt;det&gt;&lt;qnt&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> ====Additional Constraint Grammar Rules====<br /> *If I am an adjective and a quantifier but I do not agree in number with the preceding noun, I must be a quantifier<br /> **Example sentence: &quot;bilóko moké&quot; (A few things, &#039;&#039;not&#039;&#039; small things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> &quot;moké&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> ; &quot;moké&quot; adj sg REMOVE:21<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a singular adjective, I cannot be plural<br /> **Example sentence: &quot;nzelá molaí&quot; (a long road)<br /> **Old output (below) <br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> ; &quot;nzelá&quot; n cl9 pl REMOVE:25<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a plural adjective, I cannot be singular<br /> **Example sentence: &quot;nzelá milaí&quot; (long roads)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 pl<br /> ; &quot;nzelá&quot; n cl9 sg REMOVE:28<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be an ordinal and a cardinal and am modifying a plural noun, I am not an ordinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;bilóko míbalé&quot; (two things, &#039;&#039;not&#039;&#039; the second thing)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num card<br /> ; &quot;míbalé&quot; num ord REMOVE:31<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be and ordinal and a cardinal and am modifying a a singular noun, I am not a cardinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;elóko mibalé&quot; (the second thing, &#039;&#039;not&#039;&#039; two things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> ; &quot;míbalé&quot; num card REMOVE:34<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be a demonstrative nominal and a relative pronoun and am proceeded by a noun, I cannot be a demonstrative nominal.<br /> **Example sentence: &quot;Mobali óyo amonaki ngáí&quot; (The man who saw me...)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> &quot;óyo&quot; prn dem sg<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> ; &quot;óyo&quot; prn dem sg REMOVE:37<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> *Enhance existing basic verb transfer rule to work with newly implemented imperative<br /> **Example sentence: &quot;ina&quot; (sing)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#koyemba&lt;/code&gt;<br /> **New output<br /> ***Tagger:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;imp&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;yemba&lt;/code&gt;<br /> <br /> *Map Kikuyu associative to Lingala connective by removing extra agreement tags found in Kikuyu<br /> **Example sentence: &quot;wa&quot;<br /> **Old output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#ya&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;ya&lt;/code&gt;<br /> <br /> *Basic adjective transfer: keep number markings if a variable adjective in Lingala, discard otherwise<br /> **Example sentence 1: anene (big)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#monéne&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;minéne&lt;/code&gt;<br /> **Example sentence 2: mata (sour)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#bololo&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;bololo&lt;/code&gt;<br /> <br /> ===lin Transducer Evaluation===<br /> *Precision (annotated corpus): 80.53%<br /> *Recall (annotated corpus): 55.64%<br /> *Coverage (large corpus): 62.10%<br /> *Number of words (large corpus): 275,546 (actual), 304,425 (tokenized)<br /> *Number of stems in transducer: 273<br /> <br /> ===kik Transducer Evaluation===<br /> *Precision (annotated corpus):<br /> *Recall (annotated corpus): <br /> *Coverage (large corpus):<br /> *Number of words (large corpus):<br /> *Number of stems in transducer:<br /> <br /> ===lin → kik MT Evaluation===<br /> *WER (longer corpus):<br /> *PER (longer corpus):<br /> *Percentage unknown words:<br /> *Percentage known words:<br /> *Trimmed coverage over longer corpus [lin-kik.automorf.bin applied to lin.longer.txt]: 89.29%<br /> *Trimmed coverage over large corpus [lin-kik.automorf.bin applied to lin.corpus.large.txt]: 48.90%<br /> *Number of stems in longer corpus: 117 (actual) 140 (tokenized, perhaps including punctuation)<br /> *Number of stems in large corpus: 275,546 (actual), 304,425 (tokenized)<br /> <br /> ===kik → lin MT Evaluation===<br /> *WER (longer corpus): 86.32%<br /> *PER (longer corpus): 75.21%<br /> *Percentage unknown words: 39.50% [Actually probably a bit higher since some words were marked with an &quot;@&quot; and not translated while the script only looks for words marked with a &quot;*&quot;]<br /> *Percentage known words: 60.60%<br /> *Trimmed coverage over longer corpus [kik-lin.automorf.bin applied to kik.corpus.larger.txt]:<br /> *Trimmed coverage over large corpus [kik-lin.automorf.bin applied to kik.longer.txt]:<br /> *Number of stems in longer corpus:<br /> *Number of stems in large corpus:<br /> <br /> <br /> [[Category:Sp17_TranslationPairs]]<br /> [[Category:Lingala]]<br /> [[Category:Kikuyu]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Lingala_and_Kikuyu&diff=4584 2017-04-18T00:57:54Z

<p>Jmundo1: </p> <hr /> <div>Resources for machine translation between [[Lingala]] and [[Kikuyu]]. Code is stored in a [https://github.swarthmore.edu/apowell1/ling073-kik-lin GitHub Repository].<br /> ==Developed Resources==<br /> ===Parallel Corpus===<br /> <br /> A [https://github.swarthmore.edu/apowell1/ling073-lin-kik-corpus parallel corpus] created using two translations of Genesis is available on Github.<br /> <br /> ===Contrastive Grammar===<br /> Contrastive Grammar is detailed on the wiki page [[Lingala and Kikuyu/Contrastive Grammar]].<br /> <br /> ===Lexical Selection===<br /> Lexical selection is detailed on the wiki page [[Lingala and Kikuyu/Lexical Selection]].<br /> <br /> ===Structural Transfer===<br /> Structural transfer is detailed on the wiki page [[Lingala and Kikuyu/Structural Transfer]]<br /> <br /> ==Installing Unidecode For MT Evaluation==<br /> Most available Lingala text does not mark tones; however, the Lingala transducer has been configured using spellrelax to provide analyses and surface generation with tone markings included as accents on vowels. As a result, generated translations will include tone markings not present in the reference translation and artificially inflate evaluation metrics such as WER and PER. Fortunately, we can take advantage of the fact that accented characters are not present in ASCII to replace accented characters with their unaccented version on the command line.<br /> <br /> [https://pypi.python.org/pypi/Unidecode Unidecode] is a Python package that implements a function to convert Unicode strings into their closest ASCII approximation. For simple accented characters, this is precisely the mapping we need to remove the accents for evaluation against a reference translation. Unidecode can also be used on the command line with our machine translation system as follows: &lt;code&gt;echo /path/to/kik-corpus | apertium -d . kik-lin | unidecode&lt;/code&gt;. To install unidecode, it is recommended to use &lt;code&gt;pip&lt;/code&gt;, as experience suggests that &lt;code&gt;apt-get&lt;/code&gt; will not run the necessary installation scripts to make unidecode accessible in Bash.<br /> <br /> On the Ling 073 virtual machines, Pip can be installed using &lt;code&gt;sudo apt-get install python-pip&lt;/code&gt;. After installing pip, unidecode can be installed using &lt;code&gt; pip install unidecode&lt;/code&gt;. <br /> <br /> ==Lexical Selection==<br /> This section is part of the [[Lexical transfer]] assignment. In addition to adding many one-to-one mappings to our .dix file, we have included a couple of one-to-many mappings for each direction of translation. See [[Lingala and Kikuyu/Lexical Selection]] for further details.<br /> <br /> From Lingala to Kikuyu, we consider four one-to-many mappings:<br /> *The Lingala verb kokíma can be translated in English as both &quot;to run&quot; and &quot;to run away.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, hanyũk and ũr.<br /> *The Lingala verb kofánda can be translated in English as both &quot;to sit&quot; and &quot;to live&quot; (as in reside). Kikuyu has two distinct verbs corresponding to these English translations, respectively, ikar and tũũr.<br /> *The Lingala verb kobánza can be translated in English as both &quot;to think&quot; and &quot;to remember.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, ĩcir and ririkan.<br /> *The Lingala noun mabele can be translated in English as both &quot;sand&quot; and &quot;milk.&quot; Kikuyu has two distinct nouns corresponding to these English translations, respectively, thanga and ria.<br /> <br /> From Kikuyu to Lingala, we consider three one-to-many-mappings:<br /> *The Kikuyu verb igu can be translated in English as both &quot;to obey&quot; and &quot;to hear.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, kotosa and koyoka.<br /> *The Kikuyu verb rut can be translated in English as both &quot;to learn&quot; and &quot;to remove.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, koyekola and kokukula.<br /> *the Kikuyu noun ria can be translated in English as both &quot;milk&quot; and &quot;pond.&quot; Lingala has two distinct nouns corresponding to these English translations, respectively, mabele and etimá.<br /> <br /> ==Initial Evaluation==<br /> ===lin → kik evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 87.50%<br /> **PER: 87.50%<br /> **Number of words: 9<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 99.06%<br /> **PER: 94.34%<br /> **Number of words: 107<br /> **Number of unknown words: 71<br /> **Proportion of stems translated correctly:33.64%<br /> <br /> ===kik → lin evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 80%<br /> **PER: 80%<br /> **Number of words: 7 (This should really be 8, but the command we are supposed to use to count this outputs 7. The proportion remains unchanged in our case, however.)<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 98.13%<br /> **PER: 90.65%<br /> **Number of words: 106<br /> **Number of unknown words: 79 (Note: Using the commands on the wiki gives a slightly different count that using the apertium-eval-translator tool. We use the count given by the command, not the tool)<br /> **Proportion of stems translated correctly: 25.47%<br /> <br /> ==Final Evaluation==<br /> ===lin → kik Implementation Details===<br /> <br /> This direction of the pair was improved by improving the transducer to handle pronouns, proper names, and new types of morphology; improving the constraint grammar to better produce forms which agree properly; and adding transfer rules for several new input configurations including associatives and adjectives.<br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for demonstrative pronouns/determiners (proximal, distal, and anaphoric)<br /> **Example word: &quot;rĩrĩa&quot; (Class 5 singular distal demonstrative)<br /> **Old output: &lt;code&gt;^rĩrĩa/*rĩrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^rĩrĩa/pro&lt;rel&gt;&lt;cl5&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dist&gt;&lt;cl5&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for possessive pronouns/determiners<br /> **Example word: &quot;gĩakwa&quot; (Possessive pronoun for Class 7 singular object with first-person singular owner)<br /> **Old output: &lt;code&gt;^gĩakwa/*gĩakwa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gĩakwa/pro&lt;prn&gt;&lt;poss&gt;&lt;cl7&gt;&lt;sg&gt;&lt;sp1&gt;&lt;ssg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for relative pronouns<br /> **Example word: ũrĩa (Class 1 singular relative pronoun, as well as other noun classes and other types of pronouns)<br /> **Old output: &lt;code&gt;^ũrĩa/*ũrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^ũrĩa/pro&lt;prn&gt;&lt;rel&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl14&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl14&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for personal pronouns<br /> **Example word: yo (Class 3 plural personal pronoun)<br /> **Old output: &lt;code&gt;^yo/*yo$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yo/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;sg&gt;$&lt;/code&gt;<br /> <br /> *Support for proper names (proper name tag introduced, and several names added)<br /> **Example word: Kenya (Kenya)<br /> **Old output: &lt;code&gt;^Kenya/Kenya$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kenya/Kenya&lt;n&gt;&lt;prop&gt;$&lt;/code&gt;<br /> <br /> *Support for Class 16 and 17 nouns (which tend to have to do with place), and their agreement on adjectives<br /> **Example: kũndũ guothe (all places)<br /> **Old output: &lt;code&gt;^kũndũ/*kũndũ$ ^guothe/*guothe$&lt;/code&gt;<br /> **New output: &lt;code&gt;^kũndũ/ndũ&lt;n&gt;&lt;cl17&gt;&lt;sg&gt;&lt;indef&gt;$ ^guothe/othe&lt;adj&gt;&lt;cl17&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for unique reversive middle voice morphology<br /> **Example word: rũthiũrũrũkĩirie (surround, roughly)<br /> **Old output: &lt;code&gt;^makĩinũkae/*makĩinũka$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^makĩinũka/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for infinitives<br /> **Example word: gũthambĩra (to swim)<br /> **Old output: &lt;code&gt;^gũthambĩra/*gũthambĩra$ ^guothe/*guothe$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gũthambĩra/^gũthambĩra/thambĩr&lt;v&gt;&lt;inf&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for ĩ to y alternation in certain cases (implemented through twol)<br /> **Example word: yarĩ (copula form)<br /> **Old output: &lt;code&gt;^yarĩ/*yarĩ$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yarĩ/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p2&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p13&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl7&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p10&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *An attempt was made to implement object agreement morphology on verbs. Such an attempt would necessitate adding twoc rules for each noun class, and doing so made twoc take far too long to compile. This attempt was therefore abandoned, but if a workaround were to be found for twoc, the twoc rules and lexc code are commented out in their respective files.<br /> <br /> ====Additional Constraint Grammar Rules====<br /> <br /> *If I am a verb, demonstrative, or associative inflected to agree with a certain number/noun class, select me if I am preceded by an adjective preceded by a noun of that number/noun class.<br /> **Example: maaĩ maingĩ me (many waters are)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;Maaĩ&gt;&quot;<br /> &quot;aĩ&quot; n cl5 pl<br /> &quot;&lt;maingĩ&gt;&quot;<br /> &quot;ingĩ&quot; adj cl5 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl9p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl14 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl11p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl15 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl12p6 pl SELECT:27<br /> &quot;&lt;me&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:33<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/noun class, select me if I am preceded by a demonstrative preceded by a noun of that number/person/noun class.<br /> **Example: andũ acio nĩmagakorwo (those people will find)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;andũ&gt;&quot;<br /> &quot;ndũ&quot; n cl1 pl<br /> &quot;&lt;acio&gt;&quot;<br /> &quot;pro&quot; prn dem ana cl1 pl<br /> &quot;&lt;nĩmagakorwo&gt;&quot;<br /> &quot;kor&quot; v pres pass p3 cl1 pl foc SELECT:40<br /> &quot;kor&quot; v remfut pass p3 cl1 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl12p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl12p6 pl foc SELECT:40<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/person/noun class, select me if I am preceded by a number preceded by an associative preceded by a noun of that noun class.<br /> **Example: nyamũ ya mũthia nĩ (the last animal is)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;nyamũ&gt;&quot;<br /> &quot;nyamũ&quot; n cl9p10 pl<br /> &quot;nyamũ&quot; n cl9p10 sg<br /> &quot;&lt;ya&gt;&quot;<br /> &quot;a&quot; assoc cl9p10 sg SELECT:27<br /> ; &quot;a&quot; assoc cl3 pl SELECT:27<br /> ; &quot;a&quot; assoc cl9p6 sg SELECT:27<br /> &quot;&lt;mũthia&gt;&quot;<br /> &quot;mũthia&quot; num<br /> &quot;&lt;nĩ&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl9p10 pl SELECT:46<br /> &quot;rĩ&quot; cop pres p3 cl9p10 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 sg SELECT:46<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a preposition followed by a verb, remove me.<br /> **Example: na ndĩarĩ na kĩndũ (and was of nothing)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> ; &quot;na&quot; pr REMOVE:68<br /> &quot;&lt;ndĩarĩ&gt;&quot;<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p6 sg<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p10 sg<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;kĩndũ&gt;&quot;<br /> &quot;ndũ&quot; n cl7 sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> Note: None of the examples here will work, because there remains a bug in generation in the Kikuyu transducer. It cannot generate most forms it can analyze. All the forms outputted in these examples would be generated correctly if this bug is fixed, without other changes to the Kikuyu transducer.<br /> <br /> The chunker is the only part of the structural transfer system which has anything beyond the default, so only the chunker and final output is given.<br /> *Implemented transfer of adjectives<br /> **Example: ezibele monene<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^rango&lt;n&gt;&lt;cl3&gt;&lt;sg&gt;$ ^nene&lt;adj&gt;&lt;cl3&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}&lt;/code&gt;<br /> **Final output: &lt;code&gt; #rango #nene &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũrango mũnene &lt;/code&gt;<br /> <br /> *Implemented transfer of associative<br /> **Example: bato ya mokili (the people of the world)<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;pl&gt;$ ^a&lt;assoc&gt;&lt;cl1&gt;&lt;pl&gt;$ ^thĩ&lt;n&gt;&lt;cl9p10&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ a thĩ &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; andũ a thĩ &lt;/code&gt;<br /> <br /> *Implemented transfer of demonstratives<br /> **Example: moto óyo (this person)<br /> **Output of chunker: &lt;code&gt;^nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;sg&gt;$ ^pro&lt;prn&gt;&lt;dem&gt;&lt;prox&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ ũyũ&lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũndũ ũyũ &lt;/code&gt;<br /> <br /> ===kik → lin Implementation Details===<br /> This direction of the RBMT system was improved by adding items to the bilingual dictionary, adding additional morphology to the transducer, writing more constraint grammar rules, and writing additional transfer rules. Specific implementation details with examples of improved output are detailed in the sections that follow.<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for proper nouns<br /> **Example sentence: &quot;Nzambe&quot; (God)<br /> **Old output: &lt;code&gt;^Nzambe/*Nzambe$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nzambe/Nzambe&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for imperative mood<br /> **Example sentence: &quot;Kíma!&quot; (Run!)<br /> **Old output: &lt;code&gt;^Kíma/*Kíma$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kíma/kokíma&lt;v&gt;&lt;imp&gt;$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for intensifying suffixes (-a becomes -áká to denote certainty (infinitive, future) or insistence (imperative))<br /> **Example sentence 1: &quot;Azalákí kopésáká&quot; (He was really distributing...)<br /> **Old output: &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/*kopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/kopésa&lt;v&gt;&lt;inf&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 2: &quot;Akopésáká&quot; (He will really give...)<br /> **Old output: &lt;code&gt;^Akopésáká/*Akopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Akopésáká/kopésa&lt;v&gt;&lt;intns&gt;&lt;fut&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 3: &quot;Pésáká&quot; (Now do give...)<br /> **Old output: &lt;code&gt;^Pésáká/*Pésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Pésáká/kopésa&lt;v&gt;&lt;imp&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for more than 1 radical extension<br /> **Example sentence: &quot;Nabángisamákí na Pierre&quot; (I was scared by Pierre, or more literally, I was made, by Pierre, to fear something/someone)<br /> **Old output: &lt;code&gt;^Nabángisamákí/*Nabángisamákí$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nabángisamákí/kobánga&lt;v&gt;&lt;caus&gt;&lt;pass&gt;&lt;urp&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for numerals<br /> **Example sentence: &quot;zómi&quot; (ten)<br /> **Old output: &lt;code&gt;^zómi/*zómi$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^zómi/zómi&lt;num&gt;&lt;ord&gt;/zómi&lt;num&gt;&lt;card&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for quantifiers<br /> **Example sentence: &quot;mbelí nyónso&quot; (each knife)<br /> **Old output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/*nyónso$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/nyónso&lt;det&gt;&lt;qnt&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> ====Additional Constraint Grammar Rules====<br /> *If I am an adjective and a quantifier but I do not agree in number with the preceding noun, I must be a quantifier<br /> **Example sentence: &quot;bilóko moké&quot; (A few things, &#039;&#039;not&#039;&#039; small things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> &quot;moké&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> ; &quot;moké&quot; adj sg REMOVE:21<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a singular adjective, I cannot be plural<br /> **Example sentence: &quot;nzelá molaí&quot; (a long road)<br /> **Old output (below) <br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> ; &quot;nzelá&quot; n cl9 pl REMOVE:25<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a plural adjective, I cannot be singular<br /> **Example sentence: &quot;nzelá milaí&quot; (long roads)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 pl<br /> ; &quot;nzelá&quot; n cl9 sg REMOVE:28<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be an ordinal and a cardinal and am modifying a plural noun, I am not an ordinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;bilóko míbalé&quot; (two things, &#039;&#039;not&#039;&#039; the second thing)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num card<br /> ; &quot;míbalé&quot; num ord REMOVE:31<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be and ordinal and a cardinal and am modifying a a singular noun, I am not a cardinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;elóko mibalé&quot; (the second thing, &#039;&#039;not&#039;&#039; two things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> ; &quot;míbalé&quot; num card REMOVE:34<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be a demonstrative nominal and a relative pronoun and am proceeded by a noun, I cannot be a demonstrative nominal.<br /> **Example sentence: &quot;Mobali óyo amonaki ngáí&quot; (The man who saw me...)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> &quot;óyo&quot; prn dem sg<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> ; &quot;óyo&quot; prn dem sg REMOVE:37<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> *Enhance existing basic verb transfer rule to work with newly implemented imperative<br /> **Example sentence: &quot;ina&quot; (sing)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#koyemba&lt;/code&gt;<br /> **New output<br /> ***Tagger:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;imp&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;yemba&lt;/code&gt;<br /> <br /> *Map Kikuyu associative to Lingala connective by removing extra agreement tags found in Kikuyu<br /> **Example sentence: &quot;wa&quot;<br /> **Old output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#ya&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;ya&lt;/code&gt;<br /> <br /> *Basic adjective transfer: keep number markings if a variable adjective in Lingala, discard otherwise<br /> **Example sentence 1: anene (big)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#monéne&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;minéne&lt;/code&gt;<br /> **Example sentence 2: mata (sour)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#bololo&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;bololo&lt;/code&gt;<br /> <br /> ===lin Transducer Evaluation===<br /> *Precision (annotated corpus): 80.53%<br /> *Recall (annotated corpus): 55.64%<br /> *Coverage (large corpus): 62.10%<br /> *Number of words (large corpus): 275,546 (actual), 304,425 (tokenized)<br /> *Number of stems in transducer: 273<br /> <br /> ===kik Transducer Evaluation===<br /> *Precision (annotated corpus):<br /> *Recall (annotated corpus): <br /> *Coverage (large corpus):<br /> *Number of words (large corpus):<br /> *Number of stems in transducer:<br /> <br /> ===lin → kik MT Evaluation===<br /> *WER (longer corpus):<br /> *PER (longer corpus):<br /> *Percentage unknown words:<br /> *Percentage known words:<br /> *Trimmed coverage over longer corpus [lin-kik.automorf.bin applied to lin.longer.txt]: 89.29%<br /> *Trimmed coverage over large corpus [lin-kik.automorf.bin applied to lin.corpus.large.txt]: 48.90%<br /> *Number of stems in longer corpus: 117 (actual) 140 (tokenized, perhaps including punctuation)<br /> *Number of stems in large corpus: 275,546 (actual), 304,425 (tokenized)<br /> <br /> ===kik → lin MT Evaluation===<br /> *WER (longer corpus): 86.32%<br /> *PER (longer corpus): 75.21%<br /> *Percentage unknown words: 39.50% [Actually probably a bit higher since some words were marked with an &quot;@&quot; and not translated while the script only looks for words marked with a &quot;*&quot;]<br /> *Percentage known words: 60.60%<br /> *Trimmed coverage over longer corpus [kik-lin.automorf.bin applied to kik.corpus.larger.txt]:<br /> *Trimmed coverage over large corpus [kik-lin.automorf.bin applied to kik.longer.txt]:<br /> *Number of stems in longer corpus:<br /> *Number of stems in large corpus:<br /> <br /> <br /> [[Category:Sp17_TranslationPairs]]<br /> [[Category:Lingala]]<br /> [[Category:Kikuyu]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Lingala_and_Kikuyu&diff=4576 2017-04-17T22:04:48Z

<p>Jmundo1: </p> <hr /> <div>Resources for machine translation between [[Lingala]] and [[Kikuyu]]. Code is stored in a [https://github.swarthmore.edu/apowell1/ling073-kik-lin GitHub Repository].<br /> ==Developed Resources==<br /> ===Parallel Corpus===<br /> <br /> A [https://github.swarthmore.edu/apowell1/ling073-lin-kik-corpus parallel corpus] created using two translations of Genesis is available on Github.<br /> <br /> ===Contrastive Grammar===<br /> Contrastive Grammar is detailed on the wiki page [[Lingala and Kikuyu/Contrastive Grammar]].<br /> <br /> ===Lexical Selection===<br /> Lexical selection is detailed on the wiki page [[Lingala and Kikuyu/Lexical Selection]].<br /> <br /> ===Structural Transfer===<br /> Structural transfer is detailed on the wiki page [[Lingala and Kikuyu/Structural Transfer]]<br /> <br /> ==Installing Unidecode For MT Evaluation==<br /> Most available Lingala text does not mark tones; however, the Lingala transducer has been configured using spellrelax to provide analyses and surface generation with tone markings included as accents on vowels. As a result, generated translations will include tone markings not present in the reference translation and artificially inflate evaluation metrics such as WER and PER. Fortunately, we can take advantage of the fact that accented characters are not present in ASCII to replace accented characters with their unaccented version on the command line.<br /> <br /> [https://pypi.python.org/pypi/Unidecode Unidecode] is a Python package that implements a function to convert Unicode strings into their closest ASCII approximation. For simple accented characters, this is precisely the mapping we need to remove the accents for evaluation against a reference translation. Unidecode can also be used on the command line with our machine translation system as follows: &lt;code&gt;echo /path/to/kik-corpus | apertium -d . kik-lin | unidecode&lt;/code&gt;. To install unidecode, it is recommended to use &lt;code&gt;pip&lt;/code&gt;, as experience suggests that &lt;code&gt;apt-get&lt;/code&gt; will not run the necessary installation scripts to make unidecode accessible in Bash.<br /> <br /> On the Ling 073 virtual machines, Pip can be installed using &lt;code&gt;sudo apt-get install python-pip&lt;/code&gt;. After installing pip, unidecode can be installed using &lt;code&gt; pip install unidecode&lt;/code&gt;. <br /> <br /> ==Lexical Selection==<br /> This section is part of the [[Lexical transfer]] assignment. In addition to adding many one-to-one mappings to our .dix file, we have included a couple of one-to-many mappings for each direction of translation. See [[Lingala and Kikuyu/Lexical Selection]] for further details.<br /> <br /> From Lingala to Kikuyu, we consider four one-to-many mappings:<br /> *The Lingala verb kokíma can be translated in English as both &quot;to run&quot; and &quot;to run away.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, hanyũk and ũr.<br /> *The Lingala verb kofánda can be translated in English as both &quot;to sit&quot; and &quot;to live&quot; (as in reside). Kikuyu has two distinct verbs corresponding to these English translations, respectively, ikar and tũũr.<br /> *The Lingala verb kobánza can be translated in English as both &quot;to think&quot; and &quot;to remember.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, ĩcir and ririkan.<br /> *The Lingala noun mabele can be translated in English as both &quot;sand&quot; and &quot;milk.&quot; Kikuyu has two distinct nouns corresponding to these English translations, respectively, thanga and ria.<br /> <br /> From Kikuyu to Lingala, we consider three one-to-many-mappings:<br /> *The Kikuyu verb igu can be translated in English as both &quot;to obey&quot; and &quot;to hear.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, kotosa and koyoka.<br /> *The Kikuyu verb rut can be translated in English as both &quot;to learn&quot; and &quot;to remove.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, koyekola and kokukula.<br /> *the Kikuyu noun ria can be translated in English as both &quot;milk&quot; and &quot;pond.&quot; Lingala has two distinct nouns corresponding to these English translations, respectively, mabele and etimá.<br /> <br /> ==Initial Evaluation==<br /> ===lin → kik evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 87.50%<br /> **PER: 87.50%<br /> **Number of words: 9<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 99.06%<br /> **PER: 94.34%<br /> **Number of words: 107<br /> **Number of unknown words: 71<br /> **Proportion of stems translated correctly:33.64%<br /> <br /> ===kik → lin evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 80%<br /> **PER: 80%<br /> **Number of words: 7 (This should really be 8, but the command we are supposed to use to count this outputs 7. The proportion remains unchanged in our case, however.)<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 98.13%<br /> **PER: 90.65%<br /> **Number of words: 106<br /> **Number of unknown words: 79 (Note: Using the commands on the wiki gives a slightly different count that using the apertium-eval-translator tool. We use the count given by the command, not the tool)<br /> **Proportion of stems translated correctly: 25.47%<br /> <br /> ==Final Evaluation==<br /> ===lin → kik Implementation Details===<br /> <br /> This direction of the pair was improved by improving the transducer to handle pronouns, proper names, and new types of morphology; improving the constraint grammar to better produce forms which agree properly; and adding transfer rules for several new input configurations including associatives and adjectives.<br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for demonstrative pronouns/determiners (proximal, distal, and anaphoric)<br /> **Example word: &quot;rĩrĩa&quot; (Class 5 singular distal demonstrative)<br /> **Old output: &lt;code&gt;^rĩrĩa/*rĩrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^rĩrĩa/pro&lt;rel&gt;&lt;cl5&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dist&gt;&lt;cl5&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for possessive pronouns/determiners<br /> **Example word: &quot;gĩakwa&quot; (Possessive pronoun for Class 7 singular object with first-person singular owner)<br /> **Old output: &lt;code&gt;^gĩakwa/*gĩakwa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gĩakwa/pro&lt;prn&gt;&lt;poss&gt;&lt;cl7&gt;&lt;sg&gt;&lt;sp1&gt;&lt;ssg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for relative pronouns<br /> **Example word: ũrĩa (Class 1 singular relative pronoun, as well as other noun classes and other types of pronouns)<br /> **Old output: &lt;code&gt;^ũrĩa/*ũrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^ũrĩa/pro&lt;prn&gt;&lt;rel&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl14&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl14&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for personal pronouns<br /> **Example word: yo (Class 3 plural personal pronoun)<br /> **Old output: &lt;code&gt;^yo/*yo$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yo/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;sg&gt;$&lt;/code&gt;<br /> <br /> *Support for proper names (proper name tag introduced, and several names added)<br /> **Example word: Kenya (Kenya)<br /> **Old output: &lt;code&gt;^Kenya/Kenya$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kenya/Kenya&lt;n&gt;&lt;prop&gt;$&lt;/code&gt;<br /> <br /> *Support for Class 16 and 17 nouns (which tend to have to do with place), and their agreement on adjectives<br /> **Example: kũndũ guothe (all places)<br /> **Old output: &lt;code&gt;^kũndũ/*kũndũ$ ^guothe/*guothe$&lt;/code&gt;<br /> **New output: &lt;code&gt;^kũndũ/ndũ&lt;n&gt;&lt;cl17&gt;&lt;sg&gt;&lt;indef&gt;$ ^guothe/othe&lt;adj&gt;&lt;cl17&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for unique reversive middle voice morphology<br /> **Example word: rũthiũrũrũkĩirie (surround, roughly)<br /> **Old output: &lt;code&gt;^makĩinũkae/*makĩinũka$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^makĩinũka/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for ĩ to y alternation in certain cases (implemented through twol)<br /> **Example word: yarĩ (copula form)<br /> **Old output: &lt;code&gt;^yarĩ/*yarĩ$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yarĩ/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p2&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p13&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl7&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p10&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *An attempt was made to implement object agreement morphology on verbs. Such an attempt would necessitate adding twoc rules for each noun class, and doing so made twoc take far too long to compile. This attempt was therefore abandoned, but if a workaround were to be found for twoc, the twoc rules and lexc code are commented out in their respective files.<br /> <br /> ====Additional Constraint Grammar Rules====<br /> <br /> *If I am a verb, demonstrative, or associative inflected to agree with a certain number/noun class, select me if I am preceded by an adjective preceded by a noun of that number/noun class.<br /> **Example: maaĩ maingĩ me (many waters are)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;Maaĩ&gt;&quot;<br /> &quot;aĩ&quot; n cl5 pl<br /> &quot;&lt;maingĩ&gt;&quot;<br /> &quot;ingĩ&quot; adj cl5 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl9p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl14 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl11p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl15 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl12p6 pl SELECT:27<br /> &quot;&lt;me&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:33<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/noun class, select me if I am preceded by a demonstrative preceded by a noun of that number/person/noun class.<br /> **Example: andũ acio nĩmagakorwo (those people will find)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;andũ&gt;&quot;<br /> &quot;ndũ&quot; n cl1 pl<br /> &quot;&lt;acio&gt;&quot;<br /> &quot;pro&quot; prn dem ana cl1 pl<br /> &quot;&lt;nĩmagakorwo&gt;&quot;<br /> &quot;kor&quot; v pres pass p3 cl1 pl foc SELECT:40<br /> &quot;kor&quot; v remfut pass p3 cl1 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl12p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl12p6 pl foc SELECT:40<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/person/noun class, select me if I am preceded by a number preceded by an associative preceded by a noun of that noun class.<br /> **Example: nyamũ ya mũthia nĩ (the last animal is)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;nyamũ&gt;&quot;<br /> &quot;nyamũ&quot; n cl9p10 pl<br /> &quot;nyamũ&quot; n cl9p10 sg<br /> &quot;&lt;ya&gt;&quot;<br /> &quot;a&quot; assoc cl9p10 sg SELECT:27<br /> ; &quot;a&quot; assoc cl3 pl SELECT:27<br /> ; &quot;a&quot; assoc cl9p6 sg SELECT:27<br /> &quot;&lt;mũthia&gt;&quot;<br /> &quot;mũthia&quot; num<br /> &quot;&lt;nĩ&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl9p10 pl SELECT:46<br /> &quot;rĩ&quot; cop pres p3 cl9p10 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 sg SELECT:46<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a preposition followed by a verb, remove me.<br /> **Example: na ndĩarĩ na kĩndũ (and was of nothing)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> ; &quot;na&quot; pr REMOVE:68<br /> &quot;&lt;ndĩarĩ&gt;&quot;<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p6 sg<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p10 sg<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;kĩndũ&gt;&quot;<br /> &quot;ndũ&quot; n cl7 sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> Note: None of the examples here will work, because there remains a bug in generation in the Kikuyu transducer. It cannot generate most forms it can analyze. All the forms outputted in these examples would be generated correctly if this bug is fixed, without other changes to the Kikuyu transducer.<br /> <br /> The chunker is the only part of the structural transfer system which has anything beyond the default, so only the chunker and final output is given.<br /> *Implemented transfer of adjectives<br /> **Example: ezibele monene<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^rango&lt;n&gt;&lt;cl3&gt;&lt;sg&gt;$ ^nene&lt;adj&gt;&lt;cl3&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}&lt;/code&gt;<br /> **Final output: &lt;code&gt; #rango #nene &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũrango mũnene &lt;/code&gt;<br /> <br /> *Implemented transfer of associative<br /> **Example: bato ya mokili (the people of the world)<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;pl&gt;$ ^a&lt;assoc&gt;&lt;cl1&gt;&lt;pl&gt;$ ^thĩ&lt;n&gt;&lt;cl9p10&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ a thĩ &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; andũ a thĩ &lt;/code&gt;<br /> <br /> *Implemented transfer of demonstratives<br /> **Example: moto óyo (this person)<br /> **Output of chunker: &lt;code&gt;^nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;sg&gt;$ ^pro&lt;prn&gt;&lt;dem&gt;&lt;prox&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ ũyũ&lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũndũ ũyũ &lt;/code&gt;<br /> <br /> ===kik → lin Implementation Details===<br /> This direction of the RBMT system was improved by adding items to the bilingual dictionary, adding additional morphology to the transducer, writing more constraint grammar rules, and writing additional transfer rules. Specific implementation details with examples of improved output are detailed in the sections that follow.<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for proper nouns<br /> **Example sentence: &quot;Nzambe&quot; (God)<br /> **Old output: &lt;code&gt;^Nzambe/*Nzambe$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nzambe/Nzambe&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for imperative mood<br /> **Example sentence: &quot;Kíma!&quot; (Run!)<br /> **Old output: &lt;code&gt;^Kíma/*Kíma$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kíma/kokíma&lt;v&gt;&lt;imp&gt;$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for intensifying suffixes (-a becomes -áká to denote certainty (infinitive, future) or insistence (imperative))<br /> **Example sentence 1: &quot;Azalákí kopésáká&quot; (He was really distributing...)<br /> **Old output: &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/*kopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/kopésa&lt;v&gt;&lt;inf&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 2: &quot;Akopésáká&quot; (He will really give...)<br /> **Old output: &lt;code&gt;^Akopésáká/*Akopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Akopésáká/kopésa&lt;v&gt;&lt;intns&gt;&lt;fut&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 3: &quot;Pésáká&quot; (Now do give...)<br /> **Old output: &lt;code&gt;^Pésáká/*Pésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Pésáká/kopésa&lt;v&gt;&lt;imp&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for more than 1 radical extension<br /> **Example sentence: &quot;Nabángisamákí na Pierre&quot; (I was scared by Pierre, or more literally, I was made, by Pierre, to fear something/someone)<br /> **Old output: &lt;code&gt;^Nabángisamákí/*Nabángisamákí$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nabángisamákí/kobánga&lt;v&gt;&lt;caus&gt;&lt;pass&gt;&lt;urp&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for numerals<br /> **Example sentence: &quot;zómi&quot; (ten)<br /> **Old output: &lt;code&gt;^zómi/*zómi$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^zómi/zómi&lt;num&gt;&lt;ord&gt;/zómi&lt;num&gt;&lt;card&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for quantifiers<br /> **Example sentence: &quot;mbelí nyónso&quot; (each knife)<br /> **Old output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/*nyónso$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/nyónso&lt;det&gt;&lt;qnt&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> ====Additional Constraint Grammar Rules====<br /> *If I am an adjective and a quantifier but I do not agree in number with the preceding noun, I must be a quantifier<br /> **Example sentence: &quot;bilóko moké&quot; (A few things, &#039;&#039;not&#039;&#039; small things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> &quot;moké&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> ; &quot;moké&quot; adj sg REMOVE:21<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a singular adjective, I cannot be plural<br /> **Example sentence: &quot;nzelá molaí&quot; (a long road)<br /> **Old output (below) <br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> ; &quot;nzelá&quot; n cl9 pl REMOVE:25<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a plural adjective, I cannot be singular<br /> **Example sentence: &quot;nzelá milaí&quot; (long roads)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 pl<br /> ; &quot;nzelá&quot; n cl9 sg REMOVE:28<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be an ordinal and a cardinal and am modifying a plural noun, I am not an ordinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;bilóko míbalé&quot; (two things, &#039;&#039;not&#039;&#039; the second thing)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num card<br /> ; &quot;míbalé&quot; num ord REMOVE:31<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be and ordinal and a cardinal and am modifying a a singular noun, I am not a cardinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;elóko mibalé&quot; (the second thing, &#039;&#039;not&#039;&#039; two things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> ; &quot;míbalé&quot; num card REMOVE:34<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be a demonstrative nominal and a relative pronoun and am proceeded by a noun, I cannot be a demonstrative nominal.<br /> **Example sentence: &quot;Mobali óyo amonaki ngáí&quot; (The man who saw me...)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> &quot;óyo&quot; prn dem sg<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> ; &quot;óyo&quot; prn dem sg REMOVE:37<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> *Enhance existing basic verb transfer rule to work with newly implemented imperative<br /> **Example sentence: &quot;ina&quot; (sing)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#koyemba&lt;/code&gt;<br /> **New output<br /> ***Tagger:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;imp&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;yemba&lt;/code&gt;<br /> <br /> *Map Kikuyu associative to Lingala connective by removing extra agreement tags found in Kikuyu<br /> **Example sentence: &quot;wa&quot;<br /> **Old output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#ya&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;ya&lt;/code&gt;<br /> <br /> *Basic adjective transfer: keep number markings if a variable adjective in Lingala, discard otherwise<br /> **Example sentence 1: anene (big)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#monéne&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;minéne&lt;/code&gt;<br /> **Example sentence 2: mata (sour)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#bololo&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;bololo&lt;/code&gt;<br /> <br /> ===lin Transducer Evaluation===<br /> *Precision (annotated corpus): 80.53%<br /> *Recall (annotated corpus): 55.64%<br /> *Coverage (large corpus): 62.10%<br /> *Number of words (large corpus): 275,546 (actual), 304,425 (tokenized)<br /> *Number of stems in transducer: 273<br /> <br /> ===kik Transducer Evaluation===<br /> *Precision (annotated corpus):<br /> *Recall (annotated corpus): <br /> *Coverage (large corpus):<br /> *Number of words (large corpus):<br /> *Number of stems in transducer:<br /> <br /> ===lin → kik MT Evaluation===<br /> *WER (longer corpus):<br /> *PER (longer corpus):<br /> *Percentage unknown words:<br /> *Percentage known words:<br /> *Trimmed coverage over longer corpus [lin-kik.automorf.bin applied to lin.longer.txt]: 89.29%<br /> *Trimmed coverage over large corpus [lin-kik.automorf.bin applied to lin.corpus.large.txt]: 48.90%<br /> *Number of stems in longer corpus: 117 (actual) 140 (tokenized, perhaps including punctuation)<br /> *Number of stems in large corpus: 275,546 (actual), 304,425 (tokenized)<br /> <br /> ===kik → lin MT Evaluation===<br /> *WER (longer corpus): 86.32%<br /> *PER (longer corpus): 75.21%<br /> *Percentage unknown words: 39.50% [Actually probably a bit higher since some words were marked with an &quot;@&quot; and not translated while the script only looks for words marked with a &quot;*&quot;]<br /> *Percentage known words: 60.60%<br /> *Trimmed coverage over longer corpus [kik-lin.automorf.bin applied to kik.corpus.larger.txt]:<br /> *Trimmed coverage over large corpus [kik-lin.automorf.bin applied to kik.longer.txt]:<br /> *Number of stems in longer corpus:<br /> *Number of stems in large corpus:<br /> <br /> <br /> [[Category:Sp17_TranslationPairs]]<br /> [[Category:Lingala]]<br /> [[Category:Kikuyu]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Lingala_and_Kikuyu&diff=4575 2017-04-17T22:02:16Z

<p>Jmundo1: </p> <hr /> <div>Resources for machine translation between [[Lingala]] and [[Kikuyu]]. Code is stored in a [https://github.swarthmore.edu/apowell1/ling073-kik-lin GitHub Repository].<br /> ==Developed Resources==<br /> ===Parallel Corpus===<br /> <br /> A [https://github.swarthmore.edu/apowell1/ling073-lin-kik-corpus parallel corpus] created using two translations of Genesis is available on Github.<br /> <br /> ===Contrastive Grammar===<br /> Contrastive Grammar is detailed on the wiki page [[Lingala and Kikuyu/Contrastive Grammar]].<br /> <br /> ===Lexical Selection===<br /> Lexical selection is detailed on the wiki page [[Lingala and Kikuyu/Lexical Selection]].<br /> <br /> ===Structural Transfer===<br /> Structural transfer is detailed on the wiki page [[Lingala and Kikuyu/Structural Transfer]]<br /> <br /> ==Installing Unidecode For MT Evaluation==<br /> Most available Lingala text does not mark tones; however, the Lingala transducer has been configured using spellrelax to provide analyses and surface generation with tone markings included as accents on vowels. As a result, generated translations will include tone markings not present in the reference translation and artificially inflate evaluation metrics such as WER and PER. Fortunately, we can take advantage of the fact that accented characters are not present in ASCII to replace accented characters with their unaccented version on the command line.<br /> <br /> [https://pypi.python.org/pypi/Unidecode Unidecode] is a Python package that implements a function to convert Unicode strings into their closest ASCII approximation. For simple accented characters, this is precisely the mapping we need to remove the accents for evaluation against a reference translation. Unidecode can also be used on the command line with our machine translation system as follows: &lt;code&gt;echo /path/to/kik-corpus | apertium -d . kik-lin | unidecode&lt;/code&gt;. To install unidecode, it is recommended to use &lt;code&gt;pip&lt;/code&gt;, as experience suggests that &lt;code&gt;apt-get&lt;/code&gt; will not run the necessary installation scripts to make unidecode accessible in Bash.<br /> <br /> On the Ling 073 virtual machines, Pip can be installed using &lt;code&gt;sudo apt-get install python-pip&lt;/code&gt;. After installing pip, unidecode can be installed using &lt;code&gt; pip install unidecode&lt;/code&gt;. <br /> <br /> ==Lexical Selection==<br /> This section is part of the [[Lexical transfer]] assignment. In addition to adding many one-to-one mappings to our .dix file, we have included a couple of one-to-many mappings for each direction of translation. See [[Lingala and Kikuyu/Lexical Selection]] for further details.<br /> <br /> From Lingala to Kikuyu, we consider four one-to-many mappings:<br /> *The Lingala verb kokíma can be translated in English as both &quot;to run&quot; and &quot;to run away.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, hanyũk and ũr.<br /> *The Lingala verb kofánda can be translated in English as both &quot;to sit&quot; and &quot;to live&quot; (as in reside). Kikuyu has two distinct verbs corresponding to these English translations, respectively, ikar and tũũr.<br /> *The Lingala verb kobánza can be translated in English as both &quot;to think&quot; and &quot;to remember.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, ĩcir and ririkan.<br /> *The Lingala noun mabele can be translated in English as both &quot;sand&quot; and &quot;milk.&quot; Kikuyu has two distinct nouns corresponding to these English translations, respectively, thanga and ria.<br /> <br /> From Kikuyu to Lingala, we consider three one-to-many-mappings:<br /> *The Kikuyu verb igu can be translated in English as both &quot;to obey&quot; and &quot;to hear.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, kotosa and koyoka.<br /> *The Kikuyu verb rut can be translated in English as both &quot;to learn&quot; and &quot;to remove.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, koyekola and kokukula.<br /> *the Kikuyu noun ria can be translated in English as both &quot;milk&quot; and &quot;pond.&quot; Lingala has two distinct nouns corresponding to these English translations, respectively, mabele and etimá.<br /> <br /> ==Initial Evaluation==<br /> ===lin → kik evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 87.50%<br /> **PER: 87.50%<br /> **Number of words: 9<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 99.06%<br /> **PER: 94.34%<br /> **Number of words: 107<br /> **Number of unknown words: 71<br /> **Proportion of stems translated correctly:33.64%<br /> <br /> ===kik → lin evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 80%<br /> **PER: 80%<br /> **Number of words: 7 (This should really be 8, but the command we are supposed to use to count this outputs 7. The proportion remains unchanged in our case, however.)<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 98.13%<br /> **PER: 90.65%<br /> **Number of words: 106<br /> **Number of unknown words: 79 (Note: Using the commands on the wiki gives a slightly different count that using the apertium-eval-translator tool. We use the count given by the command, not the tool)<br /> **Proportion of stems translated correctly: 25.47%<br /> <br /> ==Final Evaluation==<br /> ===lin → kik Implementation Details===<br /> <br /> This direction of the pair was improved by improving the transducer to handle pronouns, proper names, and new types of morphology; improving the constraint grammar to better produce forms which agree properly; and adding transfer rules for several new input configurations including associatives and adjectives.<br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for demonstrative pronouns/determiners (proximal, distal, and anaphoric)<br /> **Example word: &quot;rĩrĩa&quot; (Class 5 singular distal demonstrative)<br /> **Old output: &lt;code&gt;^rĩrĩa/*rĩrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^rĩrĩa/pro&lt;rel&gt;&lt;cl5&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dist&gt;&lt;cl5&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for possessive pronouns/determiners<br /> **Example word: &quot;gĩakwa&quot; (Possessive pronoun for Class 7 singular object with first-person singular owner)<br /> **Old output: &lt;code&gt;^gĩakwa/*gĩakwa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gĩakwa/pro&lt;prn&gt;&lt;poss&gt;&lt;cl7&gt;&lt;sg&gt;&lt;sp1&gt;&lt;ssg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for relative pronouns<br /> **Example word: ũrĩa (Class 1 singular relative pronoun, as well as other noun classes and other types of pronouns)<br /> **Old output: &lt;code&gt;^ũrĩa/*ũrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^ũrĩa/pro&lt;prn&gt;&lt;rel&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl14&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl14&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for personal pronouns<br /> **Example word: yo (Class 3 plural personal pronoun)<br /> **Old output: &lt;code&gt;^yo/*yo$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yo/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;sg&gt;$&lt;/code&gt;<br /> <br /> *Support for proper names (proper name tag introduced, and several names added)<br /> **Example word: Kenya (Kenya)<br /> **Old output: &lt;code&gt;^Kenya/Kenya$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kenya/Kenya&lt;n&gt;&lt;prop&gt;$&lt;/code&gt;<br /> <br /> *Support for Class 16 and 17 nouns (which tend to have to do with place), and their agreement on adjectives<br /> **Example: kũndũ guothe (all places)<br /> **Old output: &lt;code&gt;^kũndũ/*kũndũ$ ^guothe/*guothe$&lt;/code&gt;<br /> **New output: &lt;code&gt;^kũndũ/ndũ&lt;n&gt;&lt;cl17&gt;&lt;sg&gt;&lt;indef&gt;$ ^guothe/othe&lt;adj&gt;&lt;cl17&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for unique reversive middle voice morphology<br /> **Example word: rũthiũrũrũkĩirie (surround, roughly)<br /> **Old output: &lt;code&gt;^makĩinũkae/*makĩinũka$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^makĩinũka/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for ĩ to y alternation in certain cases (implemented through twol)<br /> **Example word: yarĩ (copula form)<br /> **Old output: &lt;code&gt;^yarĩ/*yarĩ$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yarĩ/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p2&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p13&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl7&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p10&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *An attempt was made to implement object agreement morphology on verbs. Such an attempt would necessitate adding twoc rules for each noun class, and doing so made twoc take far too long to compile. This attempt was therefore abandoned, but if a workaround were to be found for twoc, the twoc rules and lexc code are commented out in their respective files.<br /> <br /> ====Additional Constraint Grammar Rules====<br /> <br /> *If I am a verb, demonstrative, or associative inflected to agree with a certain number/noun class, select me if I am preceded by an adjective preceded by a noun of that number/noun class.<br /> **Example: maaĩ maingĩ me (many waters are)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;Maaĩ&gt;&quot;<br /> &quot;aĩ&quot; n cl5 pl<br /> &quot;&lt;maingĩ&gt;&quot;<br /> &quot;ingĩ&quot; adj cl5 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl9p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl14 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl11p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl15 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl12p6 pl SELECT:27<br /> &quot;&lt;me&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:33<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/noun class, select me if I am preceded by a demonstrative preceded by a noun of that number/person/noun class.<br /> **Example: andũ acio nĩmagakorwo (those people will find)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;andũ&gt;&quot;<br /> &quot;ndũ&quot; n cl1 pl<br /> &quot;&lt;acio&gt;&quot;<br /> &quot;pro&quot; prn dem ana cl1 pl<br /> &quot;&lt;nĩmagakorwo&gt;&quot;<br /> &quot;kor&quot; v pres pass p3 cl1 pl foc SELECT:40<br /> &quot;kor&quot; v remfut pass p3 cl1 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl12p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl12p6 pl foc SELECT:40<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/person/noun class, select me if I am preceded by a number preceded by an associative preceded by a noun of that noun class.<br /> **Example: nyamũ ya mũthia nĩ (the last animal is)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;nyamũ&gt;&quot;<br /> &quot;nyamũ&quot; n cl9p10 pl<br /> &quot;nyamũ&quot; n cl9p10 sg<br /> &quot;&lt;ya&gt;&quot;<br /> &quot;a&quot; assoc cl9p10 sg SELECT:27<br /> ; &quot;a&quot; assoc cl3 pl SELECT:27<br /> ; &quot;a&quot; assoc cl9p6 sg SELECT:27<br /> &quot;&lt;mũthia&gt;&quot;<br /> &quot;mũthia&quot; num<br /> &quot;&lt;nĩ&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl9p10 pl SELECT:46<br /> &quot;rĩ&quot; cop pres p3 cl9p10 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 sg SELECT:46<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a preposition followed by a verb, remove me.<br /> **Example: na ndĩarĩ na kĩndũ (and was of nothing)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> ; &quot;na&quot; pr REMOVE:68<br /> &quot;&lt;ndĩarĩ&gt;&quot;<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p6 sg<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p10 sg<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;kĩndũ&gt;&quot;<br /> &quot;ndũ&quot; n cl7 sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> Note: None of the examples here will work, because there remains a bug in generation in the Kikuyu transducer. It cannot generate most forms it can analyze. All the forms outputted in these examples would be generated correctly if this bug is fixed, without other changes to the Kikuyu transducer.<br /> <br /> The chunker is the only part of the structural transfer system which has anything beyond the default, so only the chunker and final output is given.<br /> *Implemented transfer of adjectives<br /> **Example: ezibele monene<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^rango&lt;n&gt;&lt;cl3&gt;&lt;sg&gt;$ ^nene&lt;adj&gt;&lt;cl3&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}&lt;/code&gt;<br /> **<br /> **Final output: &lt;code&gt; #rango #nene &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũrango mũnene &lt;/code&gt;<br /> <br /> *Implemented transfer of associative<br /> **Example: bato ya mokili (the people of the world)<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;pl&gt;$ ^a&lt;assoc&gt;&lt;cl1&gt;&lt;pl&gt;$ ^thĩ&lt;n&gt;&lt;cl9p10&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ a thĩ &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; andũ a thĩ &lt;/code&gt;<br /> <br /> *Implemented transfer of demonstratives<br /> **Example: moto óyo (this person)<br /> **Output of chunker: &lt;code&gt;^nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;sg&gt;$ ^pro&lt;prn&gt;&lt;dem&gt;&lt;prox&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ ũyũ&lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũndũ ũyũ &lt;/code&gt;<br /> <br /> ===kik → lin Implementation Details===<br /> This direction of the RBMT system was improved by adding items to the bilingual dictionary, adding additional morphology to the transducer, writing more constraint grammar rules, and writing additional transfer rules. Specific implementation details with examples of improved output are detailed in the sections that follow.<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for proper nouns<br /> **Example sentence: &quot;Nzambe&quot; (God)<br /> **Old output: &lt;code&gt;^Nzambe/*Nzambe$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nzambe/Nzambe&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for imperative mood<br /> **Example sentence: &quot;Kíma!&quot; (Run!)<br /> **Old output: &lt;code&gt;^Kíma/*Kíma$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kíma/kokíma&lt;v&gt;&lt;imp&gt;$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for intensifying suffixes (-a becomes -áká to denote certainty (infinitive, future) or insistence (imperative))<br /> **Example sentence 1: &quot;Azalákí kopésáká&quot; (He was really distributing...)<br /> **Old output: &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/*kopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/kopésa&lt;v&gt;&lt;inf&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 2: &quot;Akopésáká&quot; (He will really give...)<br /> **Old output: &lt;code&gt;^Akopésáká/*Akopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Akopésáká/kopésa&lt;v&gt;&lt;intns&gt;&lt;fut&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 3: &quot;Pésáká&quot; (Now do give...)<br /> **Old output: &lt;code&gt;^Pésáká/*Pésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Pésáká/kopésa&lt;v&gt;&lt;imp&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for more than 1 radical extension<br /> **Example sentence: &quot;Nabángisamákí na Pierre&quot; (I was scared by Pierre, or more literally, I was made, by Pierre, to fear something/someone)<br /> **Old output: &lt;code&gt;^Nabángisamákí/*Nabángisamákí$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nabángisamákí/kobánga&lt;v&gt;&lt;caus&gt;&lt;pass&gt;&lt;urp&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for numerals<br /> **Example sentence: &quot;zómi&quot; (ten)<br /> **Old output: &lt;code&gt;^zómi/*zómi$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^zómi/zómi&lt;num&gt;&lt;ord&gt;/zómi&lt;num&gt;&lt;card&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for quantifiers<br /> **Example sentence: &quot;mbelí nyónso&quot; (each knife)<br /> **Old output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/*nyónso$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/nyónso&lt;det&gt;&lt;qnt&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> ====Additional Constraint Grammar Rules====<br /> *If I am an adjective and a quantifier but I do not agree in number with the preceding noun, I must be a quantifier<br /> **Example sentence: &quot;bilóko moké&quot; (A few things, &#039;&#039;not&#039;&#039; small things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> &quot;moké&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> ; &quot;moké&quot; adj sg REMOVE:21<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a singular adjective, I cannot be plural<br /> **Example sentence: &quot;nzelá molaí&quot; (a long road)<br /> **Old output (below) <br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> ; &quot;nzelá&quot; n cl9 pl REMOVE:25<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a plural adjective, I cannot be singular<br /> **Example sentence: &quot;nzelá milaí&quot; (long roads)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 pl<br /> ; &quot;nzelá&quot; n cl9 sg REMOVE:28<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be an ordinal and a cardinal and am modifying a plural noun, I am not an ordinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;bilóko míbalé&quot; (two things, &#039;&#039;not&#039;&#039; the second thing)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num card<br /> ; &quot;míbalé&quot; num ord REMOVE:31<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be and ordinal and a cardinal and am modifying a a singular noun, I am not a cardinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;elóko mibalé&quot; (the second thing, &#039;&#039;not&#039;&#039; two things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> ; &quot;míbalé&quot; num card REMOVE:34<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be a demonstrative nominal and a relative pronoun and am proceeded by a noun, I cannot be a demonstrative nominal.<br /> **Example sentence: &quot;Mobali óyo amonaki ngáí&quot; (The man who saw me...)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> &quot;óyo&quot; prn dem sg<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> ; &quot;óyo&quot; prn dem sg REMOVE:37<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> *Enhance existing basic verb transfer rule to work with newly implemented imperative<br /> **Example sentence: &quot;ina&quot; (sing)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#koyemba&lt;/code&gt;<br /> **New output<br /> ***Tagger:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;imp&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;yemba&lt;/code&gt;<br /> <br /> *Map Kikuyu associative to Lingala connective by removing extra agreement tags found in Kikuyu<br /> **Example sentence: &quot;wa&quot;<br /> **Old output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#ya&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;ya&lt;/code&gt;<br /> <br /> *Basic adjective transfer: keep number markings if a variable adjective in Lingala, discard otherwise<br /> **Example sentence 1: anene (big)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#monéne&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;minéne&lt;/code&gt;<br /> **Example sentence 2: mata (sour)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#bololo&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;bololo&lt;/code&gt;<br /> <br /> ===lin Transducer Evaluation===<br /> *Precision (annotated corpus): 80.53%<br /> *Recall (annotated corpus): 55.64%<br /> *Coverage (large corpus): 62.10%<br /> *Number of words (large corpus): 275,546 (actual), 304,425 (tokenized)<br /> *Number of stems in transducer: 273<br /> <br /> ===kik Transducer Evaluation===<br /> *Precision (annotated corpus):<br /> *Recall (annotated corpus): <br /> *Coverage (large corpus):<br /> *Number of words (large corpus):<br /> *Number of stems in transducer:<br /> <br /> ===lin → kik MT Evaluation===<br /> *WER (longer corpus):<br /> *PER (longer corpus):<br /> *Percentage unknown words:<br /> *Percentage known words:<br /> *Trimmed coverage over longer corpus [lin-kik.automorf.bin applied to lin.longer.txt]: 89.29%<br /> *Trimmed coverage over large corpus [lin-kik.automorf.bin applied to lin.corpus.large.txt]: 48.90%<br /> *Number of stems in longer corpus: 117 (actual) 140 (tokenized, perhaps including punctuation)<br /> *Number of stems in large corpus: 275,546 (actual), 304,425 (tokenized)<br /> <br /> ===kik → lin MT Evaluation===<br /> *WER (longer corpus): 86.32%<br /> *PER (longer corpus): 75.21%<br /> *Percentage unknown words: 39.50% [Actually probably a bit higher since some words were marked with an &quot;@&quot; and not translated while the script only looks for words marked with a &quot;*&quot;]<br /> *Percentage known words: 60.60%<br /> *Trimmed coverage over longer corpus [kik-lin.automorf.bin applied to kik.corpus.larger.txt]:<br /> *Trimmed coverage over large corpus [kik-lin.automorf.bin applied to kik.longer.txt]:<br /> *Number of stems in longer corpus:<br /> *Number of stems in large corpus:<br /> <br /> <br /> [[Category:Sp17_TranslationPairs]]<br /> [[Category:Lingala]]<br /> [[Category:Kikuyu]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Lingala_and_Kikuyu&diff=4574 2017-04-17T22:01:46Z

<p>Jmundo1: </p> <hr /> <div>Resources for machine translation between [[Lingala]] and [[Kikuyu]]. Code is stored in a [https://github.swarthmore.edu/apowell1/ling073-kik-lin GitHub Repository].<br /> ==Developed Resources==<br /> ===Parallel Corpus===<br /> <br /> A [https://github.swarthmore.edu/apowell1/ling073-lin-kik-corpus parallel corpus] created using two translations of Genesis is available on Github.<br /> <br /> ===Contrastive Grammar===<br /> Contrastive Grammar is detailed on the wiki page [[Lingala and Kikuyu/Contrastive Grammar]].<br /> <br /> ===Lexical Selection===<br /> Lexical selection is detailed on the wiki page [[Lingala and Kikuyu/Lexical Selection]].<br /> <br /> ===Structural Transfer===<br /> Structural transfer is detailed on the wiki page [[Lingala and Kikuyu/Structural Transfer]]<br /> <br /> ==Installing Unidecode For MT Evaluation==<br /> Most available Lingala text does not mark tones; however, the Lingala transducer has been configured using spellrelax to provide analyses and surface generation with tone markings included as accents on vowels. As a result, generated translations will include tone markings not present in the reference translation and artificially inflate evaluation metrics such as WER and PER. Fortunately, we can take advantage of the fact that accented characters are not present in ASCII to replace accented characters with their unaccented version on the command line.<br /> <br /> [https://pypi.python.org/pypi/Unidecode Unidecode] is a Python package that implements a function to convert Unicode strings into their closest ASCII approximation. For simple accented characters, this is precisely the mapping we need to remove the accents for evaluation against a reference translation. Unidecode can also be used on the command line with our machine translation system as follows: &lt;code&gt;echo /path/to/kik-corpus | apertium -d . kik-lin | unidecode&lt;/code&gt;. To install unidecode, it is recommended to use &lt;code&gt;pip&lt;/code&gt;, as experience suggests that &lt;code&gt;apt-get&lt;/code&gt; will not run the necessary installation scripts to make unidecode accessible in Bash.<br /> <br /> On the Ling 073 virtual machines, Pip can be installed using &lt;code&gt;sudo apt-get install python-pip&lt;/code&gt;. After installing pip, unidecode can be installed using &lt;code&gt; pip install unidecode&lt;/code&gt;. <br /> <br /> ==Lexical Selection==<br /> This section is part of the [[Lexical transfer]] assignment. In addition to adding many one-to-one mappings to our .dix file, we have included a couple of one-to-many mappings for each direction of translation. See [[Lingala and Kikuyu/Lexical Selection]] for further details.<br /> <br /> From Lingala to Kikuyu, we consider four one-to-many mappings:<br /> *The Lingala verb kokíma can be translated in English as both &quot;to run&quot; and &quot;to run away.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, hanyũk and ũr.<br /> *The Lingala verb kofánda can be translated in English as both &quot;to sit&quot; and &quot;to live&quot; (as in reside). Kikuyu has two distinct verbs corresponding to these English translations, respectively, ikar and tũũr.<br /> *The Lingala verb kobánza can be translated in English as both &quot;to think&quot; and &quot;to remember.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, ĩcir and ririkan.<br /> *The Lingala noun mabele can be translated in English as both &quot;sand&quot; and &quot;milk.&quot; Kikuyu has two distinct nouns corresponding to these English translations, respectively, thanga and ria.<br /> <br /> From Kikuyu to Lingala, we consider three one-to-many-mappings:<br /> *The Kikuyu verb igu can be translated in English as both &quot;to obey&quot; and &quot;to hear.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, kotosa and koyoka.<br /> *The Kikuyu verb rut can be translated in English as both &quot;to learn&quot; and &quot;to remove.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, koyekola and kokukula.<br /> *the Kikuyu noun ria can be translated in English as both &quot;milk&quot; and &quot;pond.&quot; Lingala has two distinct nouns corresponding to these English translations, respectively, mabele and etimá.<br /> <br /> ==Initial Evaluation==<br /> ===lin → kik evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 87.50%<br /> **PER: 87.50%<br /> **Number of words: 9<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 99.06%<br /> **PER: 94.34%<br /> **Number of words: 107<br /> **Number of unknown words: 71<br /> **Proportion of stems translated correctly:33.64%<br /> <br /> ===kik → lin evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 80%<br /> **PER: 80%<br /> **Number of words: 7 (This should really be 8, but the command we are supposed to use to count this outputs 7. The proportion remains unchanged in our case, however.)<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 98.13%<br /> **PER: 90.65%<br /> **Number of words: 106<br /> **Number of unknown words: 79 (Note: Using the commands on the wiki gives a slightly different count that using the apertium-eval-translator tool. We use the count given by the command, not the tool)<br /> **Proportion of stems translated correctly: 25.47%<br /> <br /> ==Final Evaluation==<br /> ===lin → kik Implementation Details===<br /> <br /> This direction of the pair was improved by improving the transducer to handle pronouns, proper names, and new types of morphology; improving the constraint grammar to better produce forms which agree properly; and adding transfer rules for several new input configurations including associatives and adjectives.<br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for demonstrative pronouns/determiners (proximal, distal, and anaphoric)<br /> **Example word: &quot;rĩrĩa&quot; (Class 5 singular distal demonstrative)<br /> **Old output: &lt;code&gt;^rĩrĩa/*rĩrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^rĩrĩa/pro&lt;rel&gt;&lt;cl5&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dist&gt;&lt;cl5&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for possessive pronouns/determiners<br /> **Example word: &quot;gĩakwa&quot; (Possessive pronoun for Class 7 singular object with first-person singular owner)<br /> **Old output: &lt;code&gt;^gĩakwa/*gĩakwa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gĩakwa/pro&lt;prn&gt;&lt;poss&gt;&lt;cl7&gt;&lt;sg&gt;&lt;sp1&gt;&lt;ssg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for relative pronouns<br /> **Example word: ũrĩa (Class 1 singular relative pronoun, as well as other noun classes and other types of pronouns)<br /> **Old output: &lt;code&gt;^ũrĩa/*ũrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^ũrĩa/pro&lt;prn&gt;&lt;rel&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl14&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl14&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for personal pronouns<br /> **Example word: yo (Class 3 plural personal pronoun)<br /> **Old output: &lt;code&gt;^yo/*yo$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yo/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;sg&gt;$&lt;/code&gt;<br /> <br /> *Support for proper names (proper name tag introduced, and several names added)<br /> **Example word: Kenya (Kenya)<br /> **Old output: &lt;code&gt;^Kenya/Kenya$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kenya/Kenya&lt;n&gt;&lt;prop&gt;$&lt;/code&gt;<br /> <br /> *Support for Class 16 and 17 nouns (which tend to have to do with place), and their agreement on adjectives<br /> **Example: kũndũ guothe (all places)<br /> **Old output: &lt;code&gt;^kũndũ/*kũndũ$ ^guothe/*guothe$&lt;/code&gt;<br /> **New output: &lt;code&gt;^kũndũ/ndũ&lt;n&gt;&lt;cl17&gt;&lt;sg&gt;&lt;indef&gt;$ ^guothe/othe&lt;adj&gt;&lt;cl17&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> <br /> *Support for unique reversive middle voice morphology<br /> **Example word: rũthiũrũrũkĩirie (surround, roughly)<br /> **Old output: &lt;code&gt;^makĩinũkae/*makĩinũka$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^makĩinũka/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for ĩ to y alternation in certain cases (implemented through twol)<br /> **Example word: yarĩ (copula form)<br /> **Old output: &lt;code&gt;^yarĩ/*yarĩ$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yarĩ/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p2&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p13&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl7&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p10&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *An attempt was made to implement object agreement morphology on verbs. Such an attempt would necessitate adding twoc rules for each noun class, and doing so made twoc take far too long to compile. This attempt was therefore abandoned, but if a workaround were to be found for twoc, the twoc rules and lexc code are commented out in their respective files.<br /> <br /> ====Additional Constraint Grammar Rules====<br /> <br /> *If I am a verb, demonstrative, or associative inflected to agree with a certain number/noun class, select me if I am preceded by an adjective preceded by a noun of that number/noun class.<br /> **Example: maaĩ maingĩ me (many waters are)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;Maaĩ&gt;&quot;<br /> &quot;aĩ&quot; n cl5 pl<br /> &quot;&lt;maingĩ&gt;&quot;<br /> &quot;ingĩ&quot; adj cl5 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl9p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl14 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl11p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl15 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl12p6 pl SELECT:27<br /> &quot;&lt;me&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:33<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/noun class, select me if I am preceded by a demonstrative preceded by a noun of that number/person/noun class.<br /> **Example: andũ acio nĩmagakorwo (those people will find)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;andũ&gt;&quot;<br /> &quot;ndũ&quot; n cl1 pl<br /> &quot;&lt;acio&gt;&quot;<br /> &quot;pro&quot; prn dem ana cl1 pl<br /> &quot;&lt;nĩmagakorwo&gt;&quot;<br /> &quot;kor&quot; v pres pass p3 cl1 pl foc SELECT:40<br /> &quot;kor&quot; v remfut pass p3 cl1 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl12p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl12p6 pl foc SELECT:40<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/person/noun class, select me if I am preceded by a number preceded by an associative preceded by a noun of that noun class.<br /> **Example: nyamũ ya mũthia nĩ (the last animal is)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;nyamũ&gt;&quot;<br /> &quot;nyamũ&quot; n cl9p10 pl<br /> &quot;nyamũ&quot; n cl9p10 sg<br /> &quot;&lt;ya&gt;&quot;<br /> &quot;a&quot; assoc cl9p10 sg SELECT:27<br /> ; &quot;a&quot; assoc cl3 pl SELECT:27<br /> ; &quot;a&quot; assoc cl9p6 sg SELECT:27<br /> &quot;&lt;mũthia&gt;&quot;<br /> &quot;mũthia&quot; num<br /> &quot;&lt;nĩ&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl9p10 pl SELECT:46<br /> &quot;rĩ&quot; cop pres p3 cl9p10 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 sg SELECT:46<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a preposition followed by a verb, remove me.<br /> **Example: na ndĩarĩ na kĩndũ (and was of nothing)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> ; &quot;na&quot; pr REMOVE:68<br /> &quot;&lt;ndĩarĩ&gt;&quot;<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p6 sg<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p10 sg<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;kĩndũ&gt;&quot;<br /> &quot;ndũ&quot; n cl7 sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> Note: None of the examples here will work, because there remains a bug in generation in the Kikuyu transducer. It cannot generate most forms it can analyze. All the forms outputted in these examples would be generated correctly if this bug is fixed, without other changes to the Kikuyu transducer.<br /> <br /> The chunker is the only part of the structural transfer system which has anything beyond the default, so only the chunker and final output is given.<br /> *Implemented transfer of adjectives<br /> **Example: ezibele monene<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^rango&lt;n&gt;&lt;cl3&gt;&lt;sg&gt;$ ^nene&lt;adj&gt;&lt;cl3&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}&lt;/code&gt;<br /> **<br /> **Final output: &lt;code&gt; #rango #nene &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũrango mũnene &lt;/code&gt;<br /> <br /> *Implemented transfer of associative<br /> **Example: bato ya mokili (the people of the world)<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;pl&gt;$ ^a&lt;assoc&gt;&lt;cl1&gt;&lt;pl&gt;$ ^thĩ&lt;n&gt;&lt;cl9p10&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ a thĩ &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; andũ a thĩ &lt;/code&gt;<br /> <br /> *Implemented transfer of demonstratives<br /> **Example: moto óyo (this person)<br /> **Output of chunker: &lt;code&gt;^nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;sg&gt;$ ^pro&lt;prn&gt;&lt;dem&gt;&lt;prox&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ ũyũ&lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũndũ ũyũ &lt;/code&gt;<br /> <br /> ===kik → lin Implementation Details===<br /> This direction of the RBMT system was improved by adding items to the bilingual dictionary, adding additional morphology to the transducer, writing more constraint grammar rules, and writing additional transfer rules. Specific implementation details with examples of improved output are detailed in the sections that follow.<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for proper nouns<br /> **Example sentence: &quot;Nzambe&quot; (God)<br /> **Old output: &lt;code&gt;^Nzambe/*Nzambe$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nzambe/Nzambe&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for imperative mood<br /> **Example sentence: &quot;Kíma!&quot; (Run!)<br /> **Old output: &lt;code&gt;^Kíma/*Kíma$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kíma/kokíma&lt;v&gt;&lt;imp&gt;$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for intensifying suffixes (-a becomes -áká to denote certainty (infinitive, future) or insistence (imperative))<br /> **Example sentence 1: &quot;Azalákí kopésáká&quot; (He was really distributing...)<br /> **Old output: &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/*kopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/kopésa&lt;v&gt;&lt;inf&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 2: &quot;Akopésáká&quot; (He will really give...)<br /> **Old output: &lt;code&gt;^Akopésáká/*Akopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Akopésáká/kopésa&lt;v&gt;&lt;intns&gt;&lt;fut&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 3: &quot;Pésáká&quot; (Now do give...)<br /> **Old output: &lt;code&gt;^Pésáká/*Pésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Pésáká/kopésa&lt;v&gt;&lt;imp&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for more than 1 radical extension<br /> **Example sentence: &quot;Nabángisamákí na Pierre&quot; (I was scared by Pierre, or more literally, I was made, by Pierre, to fear something/someone)<br /> **Old output: &lt;code&gt;^Nabángisamákí/*Nabángisamákí$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nabángisamákí/kobánga&lt;v&gt;&lt;caus&gt;&lt;pass&gt;&lt;urp&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for numerals<br /> **Example sentence: &quot;zómi&quot; (ten)<br /> **Old output: &lt;code&gt;^zómi/*zómi$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^zómi/zómi&lt;num&gt;&lt;ord&gt;/zómi&lt;num&gt;&lt;card&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for quantifiers<br /> **Example sentence: &quot;mbelí nyónso&quot; (each knife)<br /> **Old output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/*nyónso$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/nyónso&lt;det&gt;&lt;qnt&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> ====Additional Constraint Grammar Rules====<br /> *If I am an adjective and a quantifier but I do not agree in number with the preceding noun, I must be a quantifier<br /> **Example sentence: &quot;bilóko moké&quot; (A few things, &#039;&#039;not&#039;&#039; small things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> &quot;moké&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> ; &quot;moké&quot; adj sg REMOVE:21<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a singular adjective, I cannot be plural<br /> **Example sentence: &quot;nzelá molaí&quot; (a long road)<br /> **Old output (below) <br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> ; &quot;nzelá&quot; n cl9 pl REMOVE:25<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a plural adjective, I cannot be singular<br /> **Example sentence: &quot;nzelá milaí&quot; (long roads)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 pl<br /> ; &quot;nzelá&quot; n cl9 sg REMOVE:28<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be an ordinal and a cardinal and am modifying a plural noun, I am not an ordinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;bilóko míbalé&quot; (two things, &#039;&#039;not&#039;&#039; the second thing)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num card<br /> ; &quot;míbalé&quot; num ord REMOVE:31<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be and ordinal and a cardinal and am modifying a a singular noun, I am not a cardinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;elóko mibalé&quot; (the second thing, &#039;&#039;not&#039;&#039; two things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> ; &quot;míbalé&quot; num card REMOVE:34<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be a demonstrative nominal and a relative pronoun and am proceeded by a noun, I cannot be a demonstrative nominal.<br /> **Example sentence: &quot;Mobali óyo amonaki ngáí&quot; (The man who saw me...)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> &quot;óyo&quot; prn dem sg<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> ; &quot;óyo&quot; prn dem sg REMOVE:37<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> *Enhance existing basic verb transfer rule to work with newly implemented imperative<br /> **Example sentence: &quot;ina&quot; (sing)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#koyemba&lt;/code&gt;<br /> **New output<br /> ***Tagger:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;imp&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;yemba&lt;/code&gt;<br /> <br /> *Map Kikuyu associative to Lingala connective by removing extra agreement tags found in Kikuyu<br /> **Example sentence: &quot;wa&quot;<br /> **Old output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#ya&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;ya&lt;/code&gt;<br /> <br /> *Basic adjective transfer: keep number markings if a variable adjective in Lingala, discard otherwise<br /> **Example sentence 1: anene (big)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#monéne&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;minéne&lt;/code&gt;<br /> **Example sentence 2: mata (sour)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#bololo&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;bololo&lt;/code&gt;<br /> <br /> ===lin Transducer Evaluation===<br /> *Precision (annotated corpus): 80.53%<br /> *Recall (annotated corpus): 55.64%<br /> *Coverage (large corpus): 62.10%<br /> *Number of words (large corpus): 275,546 (actual), 304,425 (tokenized)<br /> *Number of stems in transducer: 273<br /> <br /> ===kik Transducer Evaluation===<br /> *Precision (annotated corpus):<br /> *Recall (annotated corpus): <br /> *Coverage (large corpus):<br /> *Number of words (large corpus):<br /> *Number of stems in transducer:<br /> <br /> ===lin → kik MT Evaluation===<br /> *WER (longer corpus):<br /> *PER (longer corpus):<br /> *Percentage unknown words:<br /> *Percentage known words:<br /> *Trimmed coverage over longer corpus [lin-kik.automorf.bin applied to lin.longer.txt]: 89.29%<br /> *Trimmed coverage over large corpus [lin-kik.automorf.bin applied to lin.corpus.large.txt]: 48.90%<br /> *Number of stems in longer corpus: 117 (actual) 140 (tokenized, perhaps including punctuation)<br /> *Number of stems in large corpus: 275,546 (actual), 304,425 (tokenized)<br /> <br /> ===kik → lin MT Evaluation===<br /> *WER (longer corpus): 86.32%<br /> *PER (longer corpus): 75.21%<br /> *Percentage unknown words: 39.50% [Actually probably a bit higher since some words were marked with an &quot;@&quot; and not translated while the script only looks for words marked with a &quot;*&quot;]<br /> *Percentage known words: 60.60%<br /> *Trimmed coverage over longer corpus [kik-lin.automorf.bin applied to kik.corpus.larger.txt]:<br /> *Trimmed coverage over large corpus [kik-lin.automorf.bin applied to kik.longer.txt]:<br /> *Number of stems in longer corpus:<br /> *Number of stems in large corpus:<br /> <br /> <br /> [[Category:Sp17_TranslationPairs]]<br /> [[Category:Lingala]]<br /> [[Category:Kikuyu]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Lingala_and_Kikuyu&diff=4572 2017-04-17T21:52:22Z

<p>Jmundo1: /* lin → kik Implementation Details */</p> <hr /> <div>Resources for machine translation between [[Lingala]] and [[Kikuyu]]. Code is stored in a [https://github.swarthmore.edu/apowell1/ling073-kik-lin GitHub Repository].<br /> ==Developed Resources==<br /> ===Parallel Corpus===<br /> <br /> A [https://github.swarthmore.edu/apowell1/ling073-lin-kik-corpus parallel corpus] created using two translations of Genesis is available on Github.<br /> <br /> ===Contrastive Grammar===<br /> Contrastive Grammar is detailed on the wiki page [[Lingala and Kikuyu/Contrastive Grammar]].<br /> <br /> ===Lexical Selection===<br /> Lexical selection is detailed on the wiki page [[Lingala and Kikuyu/Lexical Selection]].<br /> <br /> ===Structural Transfer===<br /> Structural transfer is detailed on the wiki page [[Lingala and Kikuyu/Structural Transfer]]<br /> <br /> ==Installing Unidecode For MT Evaluation==<br /> Most available Lingala text does not mark tones; however, the Lingala transducer has been configured using spellrelax to provide analyses and surface generation with tone markings included as accents on vowels. As a result, generated translations will include tone markings not present in the reference translation and artificially inflate evaluation metrics such as WER and PER. Fortunately, we can take advantage of the fact that accented characters are not present in ASCII to replace accented characters with their unaccented version on the command line.<br /> <br /> [https://pypi.python.org/pypi/Unidecode Unidecode] is a Python package that implements a function to convert Unicode strings into their closest ASCII approximation. For simple accented characters, this is precisely the mapping we need to remove the accents for evaluation against a reference translation. Unidecode can also be used on the command line with our machine translation system as follows: &lt;code&gt;echo /path/to/kik-corpus | apertium -d . kik-lin | unidecode&lt;/code&gt;. To install unidecode, it is recommended to use &lt;code&gt;pip&lt;/code&gt;, as experience suggests that &lt;code&gt;apt-get&lt;/code&gt; will not run the necessary installation scripts to make unidecode accessible in Bash.<br /> <br /> On the Ling 073 virtual machines, Pip can be installed using &lt;code&gt;sudo apt-get install python-pip&lt;/code&gt;. After installing pip, unidecode can be installed using &lt;code&gt; pip install unidecode&lt;/code&gt;. <br /> <br /> ==Lexical Selection==<br /> This section is part of the [[Lexical transfer]] assignment. In addition to adding many one-to-one mappings to our .dix file, we have included a couple of one-to-many mappings for each direction of translation. See [[Lingala and Kikuyu/Lexical Selection]] for further details.<br /> <br /> From Lingala to Kikuyu, we consider four one-to-many mappings:<br /> *The Lingala verb kokíma can be translated in English as both &quot;to run&quot; and &quot;to run away.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, hanyũk and ũr.<br /> *The Lingala verb kofánda can be translated in English as both &quot;to sit&quot; and &quot;to live&quot; (as in reside). Kikuyu has two distinct verbs corresponding to these English translations, respectively, ikar and tũũr.<br /> *The Lingala verb kobánza can be translated in English as both &quot;to think&quot; and &quot;to remember.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, ĩcir and ririkan.<br /> *The Lingala noun mabele can be translated in English as both &quot;sand&quot; and &quot;milk.&quot; Kikuyu has two distinct nouns corresponding to these English translations, respectively, thanga and ria.<br /> <br /> From Kikuyu to Lingala, we consider three one-to-many-mappings:<br /> *The Kikuyu verb igu can be translated in English as both &quot;to obey&quot; and &quot;to hear.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, kotosa and koyoka.<br /> *The Kikuyu verb rut can be translated in English as both &quot;to learn&quot; and &quot;to remove.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, koyekola and kokukula.<br /> *the Kikuyu noun ria can be translated in English as both &quot;milk&quot; and &quot;pond.&quot; Lingala has two distinct nouns corresponding to these English translations, respectively, mabele and etimá.<br /> <br /> ==Initial Evaluation==<br /> ===lin → kik evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 87.50%<br /> **PER: 87.50%<br /> **Number of words: 9<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 99.06%<br /> **PER: 94.34%<br /> **Number of words: 107<br /> **Number of unknown words: 71<br /> **Proportion of stems translated correctly:33.64%<br /> <br /> ===kik → lin evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 80%<br /> **PER: 80%<br /> **Number of words: 7 (This should really be 8, but the command we are supposed to use to count this outputs 7. The proportion remains unchanged in our case, however.)<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 98.13%<br /> **PER: 90.65%<br /> **Number of words: 106<br /> **Number of unknown words: 79 (Note: Using the commands on the wiki gives a slightly different count that using the apertium-eval-translator tool. We use the count given by the command, not the tool)<br /> **Proportion of stems translated correctly: 25.47%<br /> <br /> ==Final Evaluation==<br /> ===lin → kik Implementation Details===<br /> <br /> This direction of the pair was improved by improving the transducer to handle pronouns, proper names, and new types of morphology; improving the constraint grammar to better produce forms which agree properly; and adding transfer rules for several new input configurations including associatives and adjectives.<br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for demonstrative pronouns/determiners (proximal, distal, and anaphoric)<br /> **Example word: &quot;rĩrĩa&quot; (Class 5 singular distal demonstrative)<br /> **Old output: &lt;code&gt;^rĩrĩa/*rĩrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^rĩrĩa/pro&lt;rel&gt;&lt;cl5&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dist&gt;&lt;cl5&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for possessive pronouns/determiners<br /> **Example word: &quot;gĩakwa&quot; (Possessive pronoun for Class 7 singular object with first-person singular owner)<br /> **Old output: &lt;code&gt;^gĩakwa/*gĩakwa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gĩakwa/pro&lt;prn&gt;&lt;poss&gt;&lt;cl7&gt;&lt;sg&gt;&lt;sp1&gt;&lt;ssg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for relative pronouns<br /> **Example word: ũrĩa (Class 1 singular relative pronoun, as well as other noun classes and other types of pronouns)<br /> **Old output: &lt;code&gt;^ũrĩa/*ũrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^ũrĩa/pro&lt;prn&gt;&lt;rel&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl14&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl14&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for personal pronouns<br /> **Example word: yo (Class 3 plural personal pronoun)<br /> **Old output: &lt;code&gt;^yo/*yo$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yo/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;sg&gt;$&lt;/code&gt;<br /> <br /> *Support for proper names (proper name tag introduced, and several names added)<br /> **Example word: Kenya (Kenya)<br /> **Old output: &lt;code&gt;^Kenya/Kenya$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kenya/Kenya&lt;n&gt;&lt;prop&gt;$&lt;/code&gt;<br /> <br /> *Support for unique reversive middle voice morphology<br /> **Example word: rũthiũrũrũkĩirie (surround, roughly)<br /> **Old output: &lt;code&gt;^makĩinũkae/*makĩinũka$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^makĩinũka/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for ĩ to y alternation in certain cases (implemented through twol)<br /> **Example word: yarĩ (copula form)<br /> **Old output: &lt;code&gt;^yarĩ/*yarĩ$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yarĩ/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p2&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p13&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl7&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p10&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *An attempt was made to implement object agreement morphology on verbs. Such an attempt would necessitate adding twoc rules for each noun class, and doing so made twoc take far too long to compile. This attempt was therefore abandoned, but if a workaround were to be found for twoc, the twoc rules and lexc code are commented out in their respective files.<br /> <br /> ====Additional Constraint Grammar Rules====<br /> <br /> *If I am a verb, demonstrative, or associative inflected to agree with a certain number/noun class, select me if I am preceded by an adjective preceded by a noun of that number/noun class.<br /> **Example: maaĩ maingĩ me (many waters are)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;Maaĩ&gt;&quot;<br /> &quot;aĩ&quot; n cl5 pl<br /> &quot;&lt;maingĩ&gt;&quot;<br /> &quot;ingĩ&quot; adj cl5 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl9p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl14 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl11p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl15 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl12p6 pl SELECT:27<br /> &quot;&lt;me&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:33<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/noun class, select me if I am preceded by a demonstrative preceded by a noun of that number/person/noun class.<br /> **Example: andũ acio nĩmagakorwo (those people will find)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;andũ&gt;&quot;<br /> &quot;ndũ&quot; n cl1 pl<br /> &quot;&lt;acio&gt;&quot;<br /> &quot;pro&quot; prn dem ana cl1 pl<br /> &quot;&lt;nĩmagakorwo&gt;&quot;<br /> &quot;kor&quot; v pres pass p3 cl1 pl foc SELECT:40<br /> &quot;kor&quot; v remfut pass p3 cl1 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl12p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl12p6 pl foc SELECT:40<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/person/noun class, select me if I am preceded by a number preceded by an associative preceded by a noun of that noun class.<br /> **Example: nyamũ ya mũthia nĩ (the last animal is)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;nyamũ&gt;&quot;<br /> &quot;nyamũ&quot; n cl9p10 pl<br /> &quot;nyamũ&quot; n cl9p10 sg<br /> &quot;&lt;ya&gt;&quot;<br /> &quot;a&quot; assoc cl9p10 sg SELECT:27<br /> ; &quot;a&quot; assoc cl3 pl SELECT:27<br /> ; &quot;a&quot; assoc cl9p6 sg SELECT:27<br /> &quot;&lt;mũthia&gt;&quot;<br /> &quot;mũthia&quot; num<br /> &quot;&lt;nĩ&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl9p10 pl SELECT:46<br /> &quot;rĩ&quot; cop pres p3 cl9p10 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 sg SELECT:46<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a preposition followed by a verb, remove me.<br /> **Example: na ndĩarĩ na kĩndũ (and was of nothing)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> ; &quot;na&quot; pr REMOVE:68<br /> &quot;&lt;ndĩarĩ&gt;&quot;<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p6 sg<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p10 sg<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;kĩndũ&gt;&quot;<br /> &quot;ndũ&quot; n cl7 sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> Note: None of the examples here will work, because there remains a bug in generation in the Kikuyu transducer. It cannot generate most forms it can analyze. All the forms outputted in these examples would be generated correctly if this bug is fixed, without other changes to the Kikuyu transducer.<br /> <br /> The chunker is the only part of the structural transfer system which has anything beyond the default, so only the chunker and final output is given.<br /> *Implemented transfer of adjectives<br /> **Example: ezibele monene<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^rango&lt;n&gt;&lt;cl3&gt;&lt;sg&gt;$ ^nene&lt;adj&gt;&lt;cl3&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}&lt;/code&gt;<br /> **<br /> **Final output: &lt;code&gt; #rango #nene &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũrango mũnene &lt;/code&gt;<br /> <br /> *Implemented transfer of associative<br /> **Example: bato ya mokili (the people of the world)<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;pl&gt;$ ^a&lt;assoc&gt;&lt;cl1&gt;&lt;pl&gt;$ ^thĩ&lt;n&gt;&lt;cl9p10&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ a thĩ &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; andũ a thĩ &lt;/code&gt;<br /> <br /> *Implemented transfer of demonstratives<br /> **Example: moto óyo (this person)<br /> **Output of chunker: &lt;code&gt;^nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;sg&gt;$ ^pro&lt;prn&gt;&lt;dem&gt;&lt;prox&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ ũyũ&lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũndũ ũyũ &lt;/code&gt;<br /> <br /> ===kik → lin Implementation Details===<br /> This direction of the RBMT system was improved by adding items to the bilingual dictionary, adding additional morphology to the transducer, writing more constraint grammar rules, and writing additional transfer rules. Specific implementation details with examples of improved output are detailed in the sections that follow.<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for proper nouns<br /> **Example sentence: &quot;Nzambe&quot; (God)<br /> **Old output: &lt;code&gt;^Nzambe/*Nzambe$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nzambe/Nzambe&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for imperative mood<br /> **Example sentence: &quot;Kíma!&quot; (Run!)<br /> **Old output: &lt;code&gt;^Kíma/*Kíma$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kíma/kokíma&lt;v&gt;&lt;imp&gt;$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for intensifying suffixes (-a becomes -áká to denote certainty (infinitive, future) or insistence (imperative))<br /> **Example sentence 1: &quot;Azalákí kopésáká&quot; (He was really distributing...)<br /> **Old output: &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/*kopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/kopésa&lt;v&gt;&lt;inf&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 2: &quot;Akopésáká&quot; (He will really give...)<br /> **Old output: &lt;code&gt;^Akopésáká/*Akopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Akopésáká/kopésa&lt;v&gt;&lt;intns&gt;&lt;fut&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 3: &quot;Pésáká&quot; (Now do give...)<br /> **Old output: &lt;code&gt;^Pésáká/*Pésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Pésáká/kopésa&lt;v&gt;&lt;imp&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for more than 1 radical extension<br /> **Example sentence: &quot;Nabángisamákí na Pierre&quot; (I was scared by Pierre, or more literally, I was made, by Pierre, to fear something/someone)<br /> **Old output: &lt;code&gt;^Nabángisamákí/*Nabángisamákí$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nabángisamákí/kobánga&lt;v&gt;&lt;caus&gt;&lt;pass&gt;&lt;urp&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for numerals<br /> **Example sentence: &quot;zómi&quot; (ten)<br /> **Old output: &lt;code&gt;^zómi/*zómi$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^zómi/zómi&lt;num&gt;&lt;ord&gt;/zómi&lt;num&gt;&lt;card&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for quantifiers<br /> **Example sentence: &quot;mbelí nyónso&quot; (each knife)<br /> **Old output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/*nyónso$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/nyónso&lt;det&gt;&lt;qnt&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> ====Additional Constraint Grammar Rules====<br /> *If I am an adjective and a quantifier but I do not agree in number with the preceding noun, I must be a quantifier<br /> **Example sentence: &quot;bilóko moké&quot; (A few things, &#039;&#039;not&#039;&#039; small things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> &quot;moké&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> ; &quot;moké&quot; adj sg REMOVE:21<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a singular adjective, I cannot be plural<br /> **Example sentence: &quot;nzelá molaí&quot; (a long road)<br /> **Old output (below) <br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> ; &quot;nzelá&quot; n cl9 pl REMOVE:25<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a plural adjective, I cannot be singular<br /> **Example sentence: &quot;nzelá milaí&quot; (long roads)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 pl<br /> ; &quot;nzelá&quot; n cl9 sg REMOVE:28<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be an ordinal and a cardinal and am modifying a plural noun, I am not an ordinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;bilóko míbalé&quot; (two things, &#039;&#039;not&#039;&#039; the second thing)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num card<br /> ; &quot;míbalé&quot; num ord REMOVE:31<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be and ordinal and a cardinal and am modifying a a singular noun, I am not a cardinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;elóko mibalé&quot; (the second thing, &#039;&#039;not&#039;&#039; two things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> ; &quot;míbalé&quot; num card REMOVE:34<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be a demonstrative nominal and a relative pronoun and am proceeded by a noun, I cannot be a demonstrative nominal.<br /> **Example sentence: &quot;Mobali óyo amonaki ngáí&quot; (The man who saw me...)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> &quot;óyo&quot; prn dem sg<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> ; &quot;óyo&quot; prn dem sg REMOVE:37<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> *Enhance existing basic verb transfer rule to work with newly implemented imperative<br /> **Example sentence: &quot;ina&quot; (sing)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#koyemba&lt;/code&gt;<br /> **New output<br /> ***Tagger:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;imp&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;yemba&lt;/code&gt;<br /> <br /> *Map Kikuyu associative to Lingala connective by removing extra agreement tags found in Kikuyu<br /> **Example sentence: &quot;wa&quot;<br /> **Old output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#ya&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;ya&lt;/code&gt;<br /> <br /> *Basic adjective transfer: keep number markings if a variable adjective in Lingala, discard otherwise<br /> **Example sentence 1: anene (big)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#monéne&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;minéne&lt;/code&gt;<br /> **Example sentence 2: mata (sour)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#bololo&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;bololo&lt;/code&gt;<br /> <br /> ===lin Transducer Evaluation===<br /> *Precision (annotated corpus): 80.53%<br /> *Recall (annotated corpus): 55.64%<br /> *Coverage (large corpus): 62.10%<br /> *Number of words (large corpus): 275,546 (actual), 304,425 (tokenized)<br /> *Number of stems in transducer: 273<br /> <br /> ===kik Transducer Evaluation===<br /> *Precision (annotated corpus):<br /> *Recall (annotated corpus): <br /> *Coverage (large corpus):<br /> *Number of words (large corpus):<br /> *Number of stems in transducer:<br /> <br /> ===lin → kik MT Evaluation===<br /> *WER (longer corpus):<br /> *PER (longer corpus):<br /> *Percentage unknown words:<br /> *Percentage known words:<br /> *Trimmed coverage over longer corpus [lin-kik.automorf.bin applied to lin.longer.txt]: 89.29%<br /> *Trimmed coverage over large corpus [lin-kik.automorf.bin applied to lin.corpus.large.txt]: 48.90%<br /> *Number of stems in longer corpus: 117 (actual) 140 (tokenized, perhaps including punctuation)<br /> *Number of stems in large corpus: 275,546 (actual), 304,425 (tokenized)<br /> <br /> ===kik → lin MT Evaluation===<br /> *WER (longer corpus): 86.32%<br /> *PER (longer corpus): 75.21%<br /> *Percentage unknown words: 39.50% [Actually probably a bit higher since some words were marked with an &quot;@&quot; and not translated while the script only looks for words marked with a &quot;*&quot;]<br /> *Percentage known words: 60.60%<br /> *Trimmed coverage over longer corpus [kik-lin.automorf.bin applied to kik.corpus.larger.txt]:<br /> *Trimmed coverage over large corpus [kik-lin.automorf.bin applied to kik.longer.txt]:<br /> *Number of stems in longer corpus:<br /> *Number of stems in large corpus:<br /> <br /> <br /> [[Category:Sp17_TranslationPairs]]<br /> [[Category:Lingala]]<br /> [[Category:Kikuyu]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Lingala_and_Kikuyu&diff=4571 2017-04-17T21:49:13Z

<p>Jmundo1: </p> <hr /> <div>Resources for machine translation between [[Lingala]] and [[Kikuyu]]. Code is stored in a [https://github.swarthmore.edu/apowell1/ling073-kik-lin GitHub Repository].<br /> ==Developed Resources==<br /> ===Parallel Corpus===<br /> <br /> A [https://github.swarthmore.edu/apowell1/ling073-lin-kik-corpus parallel corpus] created using two translations of Genesis is available on Github.<br /> <br /> ===Contrastive Grammar===<br /> Contrastive Grammar is detailed on the wiki page [[Lingala and Kikuyu/Contrastive Grammar]].<br /> <br /> ===Lexical Selection===<br /> Lexical selection is detailed on the wiki page [[Lingala and Kikuyu/Lexical Selection]].<br /> <br /> ===Structural Transfer===<br /> Structural transfer is detailed on the wiki page [[Lingala and Kikuyu/Structural Transfer]]<br /> <br /> ==Installing Unidecode For MT Evaluation==<br /> Most available Lingala text does not mark tones; however, the Lingala transducer has been configured using spellrelax to provide analyses and surface generation with tone markings included as accents on vowels. As a result, generated translations will include tone markings not present in the reference translation and artificially inflate evaluation metrics such as WER and PER. Fortunately, we can take advantage of the fact that accented characters are not present in ASCII to replace accented characters with their unaccented version on the command line.<br /> <br /> [https://pypi.python.org/pypi/Unidecode Unidecode] is a Python package that implements a function to convert Unicode strings into their closest ASCII approximation. For simple accented characters, this is precisely the mapping we need to remove the accents for evaluation against a reference translation. Unidecode can also be used on the command line with our machine translation system as follows: &lt;code&gt;echo /path/to/kik-corpus | apertium -d . kik-lin | unidecode&lt;/code&gt;. To install unidecode, it is recommended to use &lt;code&gt;pip&lt;/code&gt;, as experience suggests that &lt;code&gt;apt-get&lt;/code&gt; will not run the necessary installation scripts to make unidecode accessible in Bash.<br /> <br /> On the Ling 073 virtual machines, Pip can be installed using &lt;code&gt;sudo apt-get install python-pip&lt;/code&gt;. After installing pip, unidecode can be installed using &lt;code&gt; pip install unidecode&lt;/code&gt;. <br /> <br /> ==Lexical Selection==<br /> This section is part of the [[Lexical transfer]] assignment. In addition to adding many one-to-one mappings to our .dix file, we have included a couple of one-to-many mappings for each direction of translation. See [[Lingala and Kikuyu/Lexical Selection]] for further details.<br /> <br /> From Lingala to Kikuyu, we consider four one-to-many mappings:<br /> *The Lingala verb kokíma can be translated in English as both &quot;to run&quot; and &quot;to run away.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, hanyũk and ũr.<br /> *The Lingala verb kofánda can be translated in English as both &quot;to sit&quot; and &quot;to live&quot; (as in reside). Kikuyu has two distinct verbs corresponding to these English translations, respectively, ikar and tũũr.<br /> *The Lingala verb kobánza can be translated in English as both &quot;to think&quot; and &quot;to remember.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, ĩcir and ririkan.<br /> *The Lingala noun mabele can be translated in English as both &quot;sand&quot; and &quot;milk.&quot; Kikuyu has two distinct nouns corresponding to these English translations, respectively, thanga and ria.<br /> <br /> From Kikuyu to Lingala, we consider three one-to-many-mappings:<br /> *The Kikuyu verb igu can be translated in English as both &quot;to obey&quot; and &quot;to hear.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, kotosa and koyoka.<br /> *The Kikuyu verb rut can be translated in English as both &quot;to learn&quot; and &quot;to remove.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, koyekola and kokukula.<br /> *the Kikuyu noun ria can be translated in English as both &quot;milk&quot; and &quot;pond.&quot; Lingala has two distinct nouns corresponding to these English translations, respectively, mabele and etimá.<br /> <br /> ==Initial Evaluation==<br /> ===lin → kik evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 87.50%<br /> **PER: 87.50%<br /> **Number of words: 9<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 99.06%<br /> **PER: 94.34%<br /> **Number of words: 107<br /> **Number of unknown words: 71<br /> **Proportion of stems translated correctly:33.64%<br /> <br /> ===kik → lin evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 80%<br /> **PER: 80%<br /> **Number of words: 7 (This should really be 8, but the command we are supposed to use to count this outputs 7. The proportion remains unchanged in our case, however.)<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 98.13%<br /> **PER: 90.65%<br /> **Number of words: 106<br /> **Number of unknown words: 79 (Note: Using the commands on the wiki gives a slightly different count that using the apertium-eval-translator tool. We use the count given by the command, not the tool)<br /> **Proportion of stems translated correctly: 25.47%<br /> <br /> ==Final Evaluation==<br /> ===lin → kik Implementation Details===<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for demonstrative pronouns/determiners (proximal, distal, and anaphoric)<br /> **Example word: &quot;rĩrĩa&quot; (Class 5 singular distal demonstrative)<br /> **Old output: &lt;code&gt;^rĩrĩa/*rĩrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^rĩrĩa/pro&lt;rel&gt;&lt;cl5&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dist&gt;&lt;cl5&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for possessive pronouns/determiners<br /> **Example word: &quot;gĩakwa&quot; (Possessive pronoun for Class 7 singular object with first-person singular owner)<br /> **Old output: &lt;code&gt;^gĩakwa/*gĩakwa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gĩakwa/pro&lt;prn&gt;&lt;poss&gt;&lt;cl7&gt;&lt;sg&gt;&lt;sp1&gt;&lt;ssg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for relative pronouns<br /> **Example word: ũrĩa (Class 1 singular relative pronoun, as well as other noun classes and other types of pronouns)<br /> **Old output: &lt;code&gt;^ũrĩa/*ũrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^ũrĩa/pro&lt;prn&gt;&lt;rel&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl14&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl14&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for personal pronouns<br /> **Example word: yo (Class 3 plural personal pronoun)<br /> **Old output: &lt;code&gt;^yo/*yo$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yo/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;sg&gt;$&lt;/code&gt;<br /> <br /> *Support for proper names (proper name tag introduced, and several names added)<br /> **Example word: Kenya (Kenya)<br /> **Old output: &lt;code&gt;^Kenya/Kenya$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kenya/Kenya&lt;n&gt;&lt;prop&gt;$&lt;/code&gt;<br /> <br /> *Support for unique reversive middle voice morphology<br /> **Example word: rũthiũrũrũkĩirie (surround, roughly)<br /> **Old output: &lt;code&gt;^makĩinũkae/*makĩinũka$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^makĩinũka/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for ĩ to y alternation in certain cases (implemented through twol)<br /> **Example word: yarĩ (copula form)<br /> **Old output: &lt;code&gt;^yarĩ/*yarĩ$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yarĩ/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p2&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p13&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl7&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p10&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *An attempt was made to implement object agreement morphology on verbs. Such an attempt would necessitate adding twoc rules for each noun class, and doing so made twoc take far too long to compile. This attempt was therefore abandoned, but if a workaround were to be found for twoc, the twoc rules and lexc code are commented out in their respective files.<br /> <br /> ====Additional Constraint Grammar Rules====<br /> <br /> *If I am a verb, demonstrative, or associative inflected to agree with a certain number/noun class, select me if I am preceded by an adjective preceded by a noun of that number/noun class.<br /> **Example: maaĩ maingĩ me (many waters are)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;Maaĩ&gt;&quot;<br /> &quot;aĩ&quot; n cl5 pl<br /> &quot;&lt;maingĩ&gt;&quot;<br /> &quot;ingĩ&quot; adj cl5 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl9p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl14 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl11p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl15 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl12p6 pl SELECT:27<br /> &quot;&lt;me&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:33<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/noun class, select me if I am preceded by a demonstrative preceded by a noun of that number/person/noun class.<br /> **Example: andũ acio nĩmagakorwo (those people will find)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;andũ&gt;&quot;<br /> &quot;ndũ&quot; n cl1 pl<br /> &quot;&lt;acio&gt;&quot;<br /> &quot;pro&quot; prn dem ana cl1 pl<br /> &quot;&lt;nĩmagakorwo&gt;&quot;<br /> &quot;kor&quot; v pres pass p3 cl1 pl foc SELECT:40<br /> &quot;kor&quot; v remfut pass p3 cl1 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl12p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl12p6 pl foc SELECT:40<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/person/noun class, select me if I am preceded by a number preceded by an associative preceded by a noun of that noun class.<br /> **Example: nyamũ ya mũthia nĩ (the last animal is)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;nyamũ&gt;&quot;<br /> &quot;nyamũ&quot; n cl9p10 pl<br /> &quot;nyamũ&quot; n cl9p10 sg<br /> &quot;&lt;ya&gt;&quot;<br /> &quot;a&quot; assoc cl9p10 sg SELECT:27<br /> ; &quot;a&quot; assoc cl3 pl SELECT:27<br /> ; &quot;a&quot; assoc cl9p6 sg SELECT:27<br /> &quot;&lt;mũthia&gt;&quot;<br /> &quot;mũthia&quot; num<br /> &quot;&lt;nĩ&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl9p10 pl SELECT:46<br /> &quot;rĩ&quot; cop pres p3 cl9p10 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 sg SELECT:46<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a preposition followed by a verb, remove me.<br /> **Example: na ndĩarĩ na kĩndũ (and was of nothing)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> ; &quot;na&quot; pr REMOVE:68<br /> &quot;&lt;ndĩarĩ&gt;&quot;<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p6 sg<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p10 sg<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;kĩndũ&gt;&quot;<br /> &quot;ndũ&quot; n cl7 sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> Note: None of the examples here will work, because there remains a bug in generation in the Kikuyu transducer. It cannot generate most forms it can analyze. All the forms outputted in these examples would be generated correctly if this bug is fixed, without other changes to the Kikuyu transducer.<br /> <br /> The chunker is the only part of the structural transfer system which has anything beyond the default, so only the chunker and final output is given.<br /> *Implemented transfer of adjectives<br /> **Example: ezibele monene<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^rango&lt;n&gt;&lt;cl3&gt;&lt;sg&gt;$ ^nene&lt;adj&gt;&lt;cl3&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}&lt;/code&gt;<br /> **<br /> **Final output: &lt;code&gt; #rango #nene &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũrango mũnene &lt;/code&gt;<br /> <br /> *Implemented transfer of associative<br /> **Example: bato ya mokili (the people of the world)<br /> **Output of chunker: &lt;code&gt;nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;pl&gt;$ ^a&lt;assoc&gt;&lt;cl1&gt;&lt;pl&gt;$ ^thĩ&lt;n&gt;&lt;cl9p10&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ a thĩ &lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; andũ a thĩ &lt;/code&gt;<br /> <br /> *Implemented transfer of demonstratives<br /> **Example: moto óyo (this person)<br /> **Output of chunker: &lt;code&gt;^nominal&lt;nominal&gt;{^ndũ&lt;n&gt;&lt;cl1&gt;&lt;sg&gt;$ ^pro&lt;prn&gt;&lt;dem&gt;&lt;prox&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> **Final output:&lt;code&gt; #ndũ ũyũ&lt;/code&gt;<br /> **Hypothetical final output if Kikuyu generation fixed: &lt;code&gt; mũndũ ũyũ &lt;/code&gt;<br /> <br /> ===kik → lin Implementation Details===<br /> This direction of the RBMT system was improved by adding items to the bilingual dictionary, adding additional morphology to the transducer, writing more constraint grammar rules, and writing additional transfer rules. Specific implementation details with examples of improved output are detailed in the sections that follow.<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for proper nouns<br /> **Example sentence: &quot;Nzambe&quot; (God)<br /> **Old output: &lt;code&gt;^Nzambe/*Nzambe$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nzambe/Nzambe&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for imperative mood<br /> **Example sentence: &quot;Kíma!&quot; (Run!)<br /> **Old output: &lt;code&gt;^Kíma/*Kíma$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kíma/kokíma&lt;v&gt;&lt;imp&gt;$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for intensifying suffixes (-a becomes -áká to denote certainty (infinitive, future) or insistence (imperative))<br /> **Example sentence 1: &quot;Azalákí kopésáká&quot; (He was really distributing...)<br /> **Old output: &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/*kopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/kopésa&lt;v&gt;&lt;inf&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 2: &quot;Akopésáká&quot; (He will really give...)<br /> **Old output: &lt;code&gt;^Akopésáká/*Akopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Akopésáká/kopésa&lt;v&gt;&lt;intns&gt;&lt;fut&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 3: &quot;Pésáká&quot; (Now do give...)<br /> **Old output: &lt;code&gt;^Pésáká/*Pésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Pésáká/kopésa&lt;v&gt;&lt;imp&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for more than 1 radical extension<br /> **Example sentence: &quot;Nabángisamákí na Pierre&quot; (I was scared by Pierre, or more literally, I was made, by Pierre, to fear something/someone)<br /> **Old output: &lt;code&gt;^Nabángisamákí/*Nabángisamákí$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nabángisamákí/kobánga&lt;v&gt;&lt;caus&gt;&lt;pass&gt;&lt;urp&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for numerals<br /> **Example sentence: &quot;zómi&quot; (ten)<br /> **Old output: &lt;code&gt;^zómi/*zómi$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^zómi/zómi&lt;num&gt;&lt;ord&gt;/zómi&lt;num&gt;&lt;card&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for quantifiers<br /> **Example sentence: &quot;mbelí nyónso&quot; (each knife)<br /> **Old output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/*nyónso$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/nyónso&lt;det&gt;&lt;qnt&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> ====Additional Constraint Grammar Rules====<br /> *If I am an adjective and a quantifier but I do not agree in number with the preceding noun, I must be a quantifier<br /> **Example sentence: &quot;bilóko moké&quot; (A few things, &#039;&#039;not&#039;&#039; small things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> &quot;moké&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> ; &quot;moké&quot; adj sg REMOVE:21<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a singular adjective, I cannot be plural<br /> **Example sentence: &quot;nzelá molaí&quot; (a long road)<br /> **Old output (below) <br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> ; &quot;nzelá&quot; n cl9 pl REMOVE:25<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a plural adjective, I cannot be singular<br /> **Example sentence: &quot;nzelá milaí&quot; (long roads)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 pl<br /> ; &quot;nzelá&quot; n cl9 sg REMOVE:28<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be an ordinal and a cardinal and am modifying a plural noun, I am not an ordinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;bilóko míbalé&quot; (two things, &#039;&#039;not&#039;&#039; the second thing)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num card<br /> ; &quot;míbalé&quot; num ord REMOVE:31<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be and ordinal and a cardinal and am modifying a a singular noun, I am not a cardinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;elóko mibalé&quot; (the second thing, &#039;&#039;not&#039;&#039; two things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> ; &quot;míbalé&quot; num card REMOVE:34<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be a demonstrative nominal and a relative pronoun and am proceeded by a noun, I cannot be a demonstrative nominal.<br /> **Example sentence: &quot;Mobali óyo amonaki ngáí&quot; (The man who saw me...)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> &quot;óyo&quot; prn dem sg<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> ; &quot;óyo&quot; prn dem sg REMOVE:37<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> *Enhance existing basic verb transfer rule to work with newly implemented imperative<br /> **Example sentence: &quot;ina&quot; (sing)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#koyemba&lt;/code&gt;<br /> **New output<br /> ***Tagger:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;imp&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;yemba&lt;/code&gt;<br /> <br /> *Map Kikuyu associative to Lingala connective by removing extra agreement tags found in Kikuyu<br /> **Example sentence: &quot;wa&quot;<br /> **Old output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#ya&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;ya&lt;/code&gt;<br /> <br /> *Basic adjective transfer: keep number markings if a variable adjective in Lingala, discard otherwise<br /> **Example sentence 1: anene (big)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#monéne&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;minéne&lt;/code&gt;<br /> **Example sentence 2: mata (sour)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#bololo&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;bololo&lt;/code&gt;<br /> <br /> ===lin Transducer Evaluation===<br /> *Precision (annotated corpus): 80.53%<br /> *Recall (annotated corpus): 55.64%<br /> *Coverage (large corpus): 62.10%<br /> *Number of words (large corpus): 275,546 (actual), 304,425 (tokenized)<br /> *Number of stems in transducer: 273<br /> <br /> ===kik Transducer Evaluation===<br /> *Precision (annotated corpus):<br /> *Recall (annotated corpus): <br /> *Coverage (large corpus):<br /> *Number of words (large corpus):<br /> *Number of stems in transducer:<br /> <br /> ===lin → kik MT Evaluation===<br /> *WER (longer corpus):<br /> *PER (longer corpus):<br /> *Percentage unknown words:<br /> *Percentage known words:<br /> *Trimmed coverage over longer corpus [lin-kik.automorf.bin applied to lin.longer.txt]: 89.29%<br /> *Trimmed coverage over large corpus [lin-kik.automorf.bin applied to lin.corpus.large.txt]: 48.90%<br /> *Number of stems in longer corpus: 117 (actual) 140 (tokenized, perhaps including punctuation)<br /> *Number of stems in large corpus: 275,546 (actual), 304,425 (tokenized)<br /> <br /> ===kik → lin MT Evaluation===<br /> *WER (longer corpus): 86.32%<br /> *PER (longer corpus): 75.21%<br /> *Percentage unknown words: 39.50% [Actually probably a bit higher since some words were marked with an &quot;@&quot; and not translated while the script only looks for words marked with a &quot;*&quot;]<br /> *Percentage known words: 60.60%<br /> *Trimmed coverage over longer corpus [kik-lin.automorf.bin applied to kik.corpus.larger.txt]:<br /> *Trimmed coverage over large corpus [kik-lin.automorf.bin applied to kik.longer.txt]:<br /> *Number of stems in longer corpus:<br /> *Number of stems in large corpus:<br /> <br /> <br /> [[Category:Sp17_TranslationPairs]]<br /> [[Category:Lingala]]<br /> [[Category:Kikuyu]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Lingala_and_Kikuyu&diff=4570 2017-04-17T21:09:53Z

<p>Jmundo1: </p> <hr /> <div>Resources for machine translation between [[Lingala]] and [[Kikuyu]]. Code is stored in a [https://github.swarthmore.edu/apowell1/ling073-kik-lin GitHub Repository].<br /> ==Developed Resources==<br /> ===Parallel Corpus===<br /> <br /> A [https://github.swarthmore.edu/apowell1/ling073-lin-kik-corpus parallel corpus] created using two translations of Genesis is available on Github.<br /> <br /> ===Contrastive Grammar===<br /> Contrastive Grammar is detailed on the wiki page [[Lingala and Kikuyu/Contrastive Grammar]].<br /> <br /> ===Lexical Selection===<br /> Lexical selection is detailed on the wiki page [[Lingala and Kikuyu/Lexical Selection]].<br /> <br /> ===Structural Transfer===<br /> Structural transfer is detailed on the wiki page [[Lingala and Kikuyu/Structural Transfer]]<br /> <br /> ==Installing Unidecode For MT Evaluation==<br /> Most available Lingala text does not mark tones; however, the Lingala transducer has been configured using spellrelax to provide analyses and surface generation with tone markings included as accents on vowels. As a result, generated translations will include tone markings not present in the reference translation and artificially inflate evaluation metrics such as WER and PER. Fortunately, we can take advantage of the fact that accented characters are not present in ASCII to replace accented characters with their unaccented version on the command line.<br /> <br /> [https://pypi.python.org/pypi/Unidecode Unidecode] is a Python package that implements a function to convert Unicode strings into their closest ASCII approximation. For simple accented characters, this is precisely the mapping we need to remove the accents for evaluation against a reference translation. Unidecode can also be used on the command line with our machine translation system as follows: &lt;code&gt;echo /path/to/kik-corpus | apertium -d . kik-lin | unidecode&lt;/code&gt;. To install unidecode, it is recommended to use &lt;code&gt;pip&lt;/code&gt;, as experience suggests that &lt;code&gt;apt-get&lt;/code&gt; will not run the necessary installation scripts to make unidecode accessible in Bash.<br /> <br /> On the Ling 073 virtual machines, Pip can be installed using &lt;code&gt;sudo apt-get install python-pip&lt;/code&gt;. After installing pip, unidecode can be installed using &lt;code&gt; pip install unidecode&lt;/code&gt;. <br /> <br /> ==Lexical Selection==<br /> This section is part of the [[Lexical transfer]] assignment. In addition to adding many one-to-one mappings to our .dix file, we have included a couple of one-to-many mappings for each direction of translation. See [[Lingala and Kikuyu/Lexical Selection]] for further details.<br /> <br /> From Lingala to Kikuyu, we consider four one-to-many mappings:<br /> *The Lingala verb kokíma can be translated in English as both &quot;to run&quot; and &quot;to run away.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, hanyũk and ũr.<br /> *The Lingala verb kofánda can be translated in English as both &quot;to sit&quot; and &quot;to live&quot; (as in reside). Kikuyu has two distinct verbs corresponding to these English translations, respectively, ikar and tũũr.<br /> *The Lingala verb kobánza can be translated in English as both &quot;to think&quot; and &quot;to remember.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, ĩcir and ririkan.<br /> *The Lingala noun mabele can be translated in English as both &quot;sand&quot; and &quot;milk.&quot; Kikuyu has two distinct nouns corresponding to these English translations, respectively, thanga and ria.<br /> <br /> From Kikuyu to Lingala, we consider three one-to-many-mappings:<br /> *The Kikuyu verb igu can be translated in English as both &quot;to obey&quot; and &quot;to hear.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, kotosa and koyoka.<br /> *The Kikuyu verb rut can be translated in English as both &quot;to learn&quot; and &quot;to remove.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, koyekola and kokukula.<br /> *the Kikuyu noun ria can be translated in English as both &quot;milk&quot; and &quot;pond.&quot; Lingala has two distinct nouns corresponding to these English translations, respectively, mabele and etimá.<br /> <br /> ==Initial Evaluation==<br /> ===lin → kik evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 87.50%<br /> **PER: 87.50%<br /> **Number of words: 9<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 99.06%<br /> **PER: 94.34%<br /> **Number of words: 107<br /> **Number of unknown words: 71<br /> **Proportion of stems translated correctly:33.64%<br /> <br /> ===kik → lin evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 80%<br /> **PER: 80%<br /> **Number of words: 7 (This should really be 8, but the command we are supposed to use to count this outputs 7. The proportion remains unchanged in our case, however.)<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 98.13%<br /> **PER: 90.65%<br /> **Number of words: 106<br /> **Number of unknown words: 79 (Note: Using the commands on the wiki gives a slightly different count that using the apertium-eval-translator tool. We use the count given by the command, not the tool)<br /> **Proportion of stems translated correctly: 25.47%<br /> <br /> ==Final Evaluation==<br /> ===lin → kik Implementation Details===<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for demonstrative pronouns/determiners (proximal, distal, and anaphoric)<br /> **Example word: &quot;rĩrĩa&quot; (Class 5 singular distal demonstrative)<br /> **Old output: &lt;code&gt;^rĩrĩa/*rĩrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^rĩrĩa/pro&lt;rel&gt;&lt;cl5&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dist&gt;&lt;cl5&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for possessive pronouns/determiners<br /> **Example word: &quot;gĩakwa&quot; (Possessive pronoun for Class 7 singular object with first-person singular owner)<br /> **Old output: &lt;code&gt;^gĩakwa/*gĩakwa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gĩakwa/pro&lt;prn&gt;&lt;poss&gt;&lt;cl7&gt;&lt;sg&gt;&lt;sp1&gt;&lt;ssg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for relative pronouns<br /> **Example word: ũrĩa (Class 1 singular relative pronoun, as well as other noun classes and other types of pronouns)<br /> **Old output: &lt;code&gt;^ũrĩa/*ũrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^ũrĩa/pro&lt;prn&gt;&lt;rel&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl14&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl14&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for personal pronouns<br /> **Example word: yo (Class 3 plural personal pronoun)<br /> **Old output: &lt;code&gt;^yo/*yo$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yo/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;sg&gt;$&lt;/code&gt;<br /> <br /> *Support for proper names (proper name tag introduced, and several names added)<br /> **Example word: Kenya (Kenya)<br /> **Old output: &lt;code&gt;^Kenya/Kenya$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kenya/Kenya&lt;n&gt;&lt;prop&gt;$&lt;/code&gt;<br /> <br /> *Support for unique reversive middle voice morphology<br /> **Example word: rũthiũrũrũkĩirie (surround, roughly)<br /> **Old output: &lt;code&gt;^makĩinũkae/*makĩinũka$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^makĩinũka/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for ĩ to y alternation in certain cases (implemented through twol)<br /> **Example word: yarĩ (copula form)<br /> **Old output: &lt;code&gt;^yarĩ/*yarĩ$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yarĩ/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p2&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p13&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl7&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p10&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *An attempt was made to implement object agreement morphology on verbs. Such an attempt would necessitate adding twoc rules for each noun class, and doing so made twoc take far too long to compile. This attempt was therefore abandoned, but if a workaround were to be found for twoc, the twoc rules and lexc code are commented out in their respective files.<br /> <br /> ====Additional Constraint Grammar Rules====<br /> <br /> *If I am a verb, demonstrative, or associative inflected to agree with a certain number/noun class, select me if I am preceded by an adjective preceded by a noun of that number/noun class.<br /> **Example: maaĩ maingĩ me (many waters are)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;Maaĩ&gt;&quot;<br /> &quot;aĩ&quot; n cl5 pl<br /> &quot;&lt;maingĩ&gt;&quot;<br /> &quot;ingĩ&quot; adj cl5 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl9p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl14 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl11p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl15 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl12p6 pl SELECT:27<br /> &quot;&lt;me&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:33<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/noun class, select me if I am preceded by a demonstrative preceded by a noun of that number/person/noun class.<br /> **Example: andũ acio nĩmagakorwo (those people will find)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;andũ&gt;&quot;<br /> &quot;ndũ&quot; n cl1 pl<br /> &quot;&lt;acio&gt;&quot;<br /> &quot;pro&quot; prn dem ana cl1 pl<br /> &quot;&lt;nĩmagakorwo&gt;&quot;<br /> &quot;kor&quot; v pres pass p3 cl1 pl foc SELECT:40<br /> &quot;kor&quot; v remfut pass p3 cl1 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl12p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl12p6 pl foc SELECT:40<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/person/noun class, select me if I am preceded by a number preceded by an associative preceded by a noun of that noun class.<br /> **Example: nyamũ ya mũthia nĩ (the last animal is)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;nyamũ&gt;&quot;<br /> &quot;nyamũ&quot; n cl9p10 pl<br /> &quot;nyamũ&quot; n cl9p10 sg<br /> &quot;&lt;ya&gt;&quot;<br /> &quot;a&quot; assoc cl9p10 sg SELECT:27<br /> ; &quot;a&quot; assoc cl3 pl SELECT:27<br /> ; &quot;a&quot; assoc cl9p6 sg SELECT:27<br /> &quot;&lt;mũthia&gt;&quot;<br /> &quot;mũthia&quot; num<br /> &quot;&lt;nĩ&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl9p10 pl SELECT:46<br /> &quot;rĩ&quot; cop pres p3 cl9p10 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 sg SELECT:46<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a preposition followed by a verb, remove me.<br /> **Example: na ndĩarĩ na kĩndũ (and was of nothing)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> ; &quot;na&quot; pr REMOVE:68<br /> &quot;&lt;ndĩarĩ&gt;&quot;<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p6 sg<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p10 sg<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;kĩndũ&gt;&quot;<br /> &quot;ndũ&quot; n cl7 sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> <br /> <br /> ===kik → lin Implementation Details===<br /> This direction of the RBMT system was improved by adding items to the bilingual dictionary, adding additional morphology to the transducer, writing more constraint grammar rules, and writing additional transfer rules. Specific implementation details with examples of improved output are detailed in the sections that follow.<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for proper nouns<br /> **Example sentence: &quot;Nzambe&quot; (God)<br /> **Old output: &lt;code&gt;^Nzambe/*Nzambe$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nzambe/Nzambe&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for imperative mood<br /> **Example sentence: &quot;Kíma!&quot; (Run!)<br /> **Old output: &lt;code&gt;^Kíma/*Kíma$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kíma/kokíma&lt;v&gt;&lt;imp&gt;$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for intensifying suffixes (-a becomes -áká to denote certainty (infinitive, future) or insistence (imperative))<br /> **Example sentence 1: &quot;Azalákí kopésáká&quot; (He was really distributing...)<br /> **Old output: &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/*kopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/kopésa&lt;v&gt;&lt;inf&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 2: &quot;Akopésáká&quot; (He will really give...)<br /> **Old output: &lt;code&gt;^Akopésáká/*Akopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Akopésáká/kopésa&lt;v&gt;&lt;intns&gt;&lt;fut&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 3: &quot;Pésáká&quot; (Now do give...)<br /> **Old output: &lt;code&gt;^Pésáká/*Pésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Pésáká/kopésa&lt;v&gt;&lt;imp&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for more than 1 radical extension<br /> **Example sentence: &quot;Nabángisamákí na Pierre&quot; (I was scared by Pierre, or more literally, I was made, by Pierre, to fear something/someone)<br /> **Old output: &lt;code&gt;^Nabángisamákí/*Nabángisamákí$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nabángisamákí/kobánga&lt;v&gt;&lt;caus&gt;&lt;pass&gt;&lt;urp&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for numerals<br /> **Example sentence: &quot;zómi&quot; (ten)<br /> **Old output: &lt;code&gt;^zómi/*zómi$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^zómi/zómi&lt;num&gt;&lt;ord&gt;/zómi&lt;num&gt;&lt;card&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for quantifiers<br /> **Example sentence: &quot;mbelí nyónso&quot; (each knife)<br /> **Old output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/*nyónso$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/nyónso&lt;det&gt;&lt;qnt&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> ====Additional Constraint Grammar Rules====<br /> *If I am an adjective and a quantifier but I do not agree in number with the preceding noun, I must be a quantifier<br /> **Example sentence: &quot;bilóko moké&quot; (A few things, &#039;&#039;not&#039;&#039; small things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> &quot;moké&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> ; &quot;moké&quot; adj sg REMOVE:21<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a singular adjective, I cannot be plural<br /> **Example sentence: &quot;nzelá molaí&quot; (a long road)<br /> **Old output (below) <br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> ; &quot;nzelá&quot; n cl9 pl REMOVE:25<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a plural adjective, I cannot be singular<br /> **Example sentence: &quot;nzelá milaí&quot; (long roads)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 pl<br /> ; &quot;nzelá&quot; n cl9 sg REMOVE:28<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be an ordinal and a cardinal and am modifying a plural noun, I am not an ordinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;bilóko míbalé&quot; (two things, &#039;&#039;not&#039;&#039; the second thing)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num card<br /> ; &quot;míbalé&quot; num ord REMOVE:31<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be and ordinal and a cardinal and am modifying a a singular noun, I am not a cardinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;elóko mibalé&quot; (the second thing, &#039;&#039;not&#039;&#039; two things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> ; &quot;míbalé&quot; num card REMOVE:34<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be a demonstrative nominal and a relative pronoun and am proceeded by a noun, I cannot be a demonstrative nominal.<br /> **Example sentence: &quot;Mobali óyo amonaki ngáí&quot; (The man who saw me...)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> &quot;óyo&quot; prn dem sg<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> ; &quot;óyo&quot; prn dem sg REMOVE:37<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> *Enhance existing basic verb transfer rule to work with newly implemented imperative<br /> **Example sentence: &quot;ina&quot; (sing)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#koyemba&lt;/code&gt;<br /> **New output<br /> ***Tagger:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;imp&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;yemba&lt;/code&gt;<br /> <br /> *Map Kikuyu associative to Lingala connective by removing extra agreement tags found in Kikuyu<br /> **Example sentence: &quot;wa&quot;<br /> **Old output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#ya&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;ya&lt;/code&gt;<br /> <br /> *Basic adjective transfer: keep number markings if a variable adjective in Lingala, discard otherwise<br /> **Example sentence 1: anene (big)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#monéne&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;minéne&lt;/code&gt;<br /> **Example sentence 2: mata (sour)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#bololo&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;bololo&lt;/code&gt;<br /> <br /> ===lin Transducer Evaluation===<br /> *Precision (annotated corpus): 80.53%<br /> *Recall (annotated corpus): 55.64%<br /> *Coverage (large corpus): 62.10%<br /> *Number of words (large corpus): 275,546 (actual), 304,425 (tokenized)<br /> *Number of stems in transducer: 273<br /> <br /> ===kik Transducer Evaluation===<br /> *Precision (annotated corpus):<br /> *Recall (annotated corpus): <br /> *Coverage (large corpus):<br /> *Number of words (large corpus):<br /> *Number of stems in transducer:<br /> <br /> ===lin → kik MT Evaluation===<br /> *WER (longer corpus):<br /> *PER (longer corpus):<br /> *Percentage unknown words:<br /> *Percentage known words:<br /> *Trimmed coverage over longer corpus [lin-kik.automorf.bin applied to lin.longer.txt]: 89.29%<br /> *Trimmed coverage over large corpus [lin-kik.automorf.bin applied to lin.corpus.large.txt]: 48.90%<br /> *Number of stems in longer corpus: 117 (actual) 140 (tokenized, perhaps including punctuation)<br /> *Number of stems in large corpus: 275,546 (actual), 304,425 (tokenized)<br /> <br /> ===kik → lin MT Evaluation===<br /> *WER (longer corpus): 86.32%<br /> *PER (longer corpus): 75.21%<br /> *Percentage unknown words: 39.50% [Actually probably a bit higher since some words were marked with an &quot;@&quot; and not translated while the script only looks for words marked with a &quot;*&quot;]<br /> *Percentage known words: 60.60%<br /> *Trimmed coverage over longer corpus [kik-lin.automorf.bin applied to kik.corpus.larger.txt]:<br /> *Trimmed coverage over large corpus [kik-lin.automorf.bin applied to kik.longer.txt]:<br /> *Number of stems in longer corpus:<br /> *Number of stems in large corpus:<br /> <br /> <br /> [[Category:Sp17_TranslationPairs]]<br /> [[Category:Lingala]]<br /> [[Category:Kikuyu]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Lingala_and_Kikuyu&diff=4569 2017-04-17T21:08:12Z

<p>Jmundo1: </p> <hr /> <div>Resources for machine translation between [[Lingala]] and [[Kikuyu]]. Code is stored in a [https://github.swarthmore.edu/apowell1/ling073-kik-lin GitHub Repository].<br /> ==Developed Resources==<br /> ===Parallel Corpus===<br /> <br /> A [https://github.swarthmore.edu/apowell1/ling073-lin-kik-corpus parallel corpus] created using two translations of Genesis is available on Github.<br /> <br /> ===Contrastive Grammar===<br /> Contrastive Grammar is detailed on the wiki page [[Lingala and Kikuyu/Contrastive Grammar]].<br /> <br /> ===Lexical Selection===<br /> Lexical selection is detailed on the wiki page [[Lingala and Kikuyu/Lexical Selection]].<br /> <br /> ===Structural Transfer===<br /> Structural transfer is detailed on the wiki page [[Lingala and Kikuyu/Structural Transfer]]<br /> <br /> ==Installing Unidecode For MT Evaluation==<br /> Most available Lingala text does not mark tones; however, the Lingala transducer has been configured using spellrelax to provide analyses and surface generation with tone markings included as accents on vowels. As a result, generated translations will include tone markings not present in the reference translation and artificially inflate evaluation metrics such as WER and PER. Fortunately, we can take advantage of the fact that accented characters are not present in ASCII to replace accented characters with their unaccented version on the command line.<br /> <br /> [https://pypi.python.org/pypi/Unidecode Unidecode] is a Python package that implements a function to convert Unicode strings into their closest ASCII approximation. For simple accented characters, this is precisely the mapping we need to remove the accents for evaluation against a reference translation. Unidecode can also be used on the command line with our machine translation system as follows: &lt;code&gt;echo /path/to/kik-corpus | apertium -d . kik-lin | unidecode&lt;/code&gt;. To install unidecode, it is recommended to use &lt;code&gt;pip&lt;/code&gt;, as experience suggests that &lt;code&gt;apt-get&lt;/code&gt; will not run the necessary installation scripts to make unidecode accessible in Bash.<br /> <br /> On the Ling 073 virtual machines, Pip can be installed using &lt;code&gt;sudo apt-get install python-pip&lt;/code&gt;. After installing pip, unidecode can be installed using &lt;code&gt; pip install unidecode&lt;/code&gt;. <br /> <br /> ==Lexical Selection==<br /> This section is part of the [[Lexical transfer]] assignment. In addition to adding many one-to-one mappings to our .dix file, we have included a couple of one-to-many mappings for each direction of translation. See [[Lingala and Kikuyu/Lexical Selection]] for further details.<br /> <br /> From Lingala to Kikuyu, we consider four one-to-many mappings:<br /> *The Lingala verb kokíma can be translated in English as both &quot;to run&quot; and &quot;to run away.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, hanyũk and ũr.<br /> *The Lingala verb kofánda can be translated in English as both &quot;to sit&quot; and &quot;to live&quot; (as in reside). Kikuyu has two distinct verbs corresponding to these English translations, respectively, ikar and tũũr.<br /> *The Lingala verb kobánza can be translated in English as both &quot;to think&quot; and &quot;to remember.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, ĩcir and ririkan.<br /> *The Lingala noun mabele can be translated in English as both &quot;sand&quot; and &quot;milk.&quot; Kikuyu has two distinct nouns corresponding to these English translations, respectively, thanga and ria.<br /> <br /> From Kikuyu to Lingala, we consider three one-to-many-mappings:<br /> *The Kikuyu verb igu can be translated in English as both &quot;to obey&quot; and &quot;to hear.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, kotosa and koyoka.<br /> *The Kikuyu verb rut can be translated in English as both &quot;to learn&quot; and &quot;to remove.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, koyekola and kokukula.<br /> *the Kikuyu noun ria can be translated in English as both &quot;milk&quot; and &quot;pond.&quot; Lingala has two distinct nouns corresponding to these English translations, respectively, mabele and etimá.<br /> <br /> ==Initial Evaluation==<br /> ===lin → kik evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 87.50%<br /> **PER: 87.50%<br /> **Number of words: 9<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 99.06%<br /> **PER: 94.34%<br /> **Number of words: 107<br /> **Number of unknown words: 71<br /> **Proportion of stems translated correctly:33.64%<br /> <br /> ===kik → lin evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 80%<br /> **PER: 80%<br /> **Number of words: 7 (This should really be 8, but the command we are supposed to use to count this outputs 7. The proportion remains unchanged in our case, however.)<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 98.13%<br /> **PER: 90.65%<br /> **Number of words: 106<br /> **Number of unknown words: 79 (Note: Using the commands on the wiki gives a slightly different count that using the apertium-eval-translator tool. We use the count given by the command, not the tool)<br /> **Proportion of stems translated correctly: 25.47%<br /> <br /> ==Final Evaluation==<br /> ===lin → kik Implementation Details===<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for demonstrative pronouns/determiners (proximal, distal, and anaphoric)<br /> **Example word: &quot;rĩrĩa&quot; (Class 5 singular distal demonstrative)<br /> **Old output: &lt;code&gt;^rĩrĩa/*rĩrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^rĩrĩa/pro&lt;rel&gt;&lt;cl5&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dist&gt;&lt;cl5&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for possessive pronouns/determiners<br /> **Example word: &quot;gĩakwa&quot; (Possessive pronoun for Class 7 singular object with first-person singular owner)<br /> **Old output: &lt;code&gt;^gĩakwa/*gĩakwa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gĩakwa/pro&lt;prn&gt;&lt;poss&gt;&lt;cl7&gt;&lt;sg&gt;&lt;sp1&gt;&lt;ssg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for relative pronouns<br /> **Example word: ũrĩa (Class 1 singular relative pronoun, as well as other noun classes and other types of pronouns)<br /> **Old output: &lt;code&gt;^ũrĩa/*ũrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^ũrĩa/pro&lt;prn&gt;&lt;rel&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl14&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl14&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for personal pronouns<br /> **Example word: yo (Class 3 plural personal pronoun)<br /> **Old output: &lt;code&gt;^yo/*yo$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yo/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;sg&gt;$&lt;/code&gt;<br /> <br /> *Support for proper names (proper name tag introduced, and several names added)<br /> **Example word: Kenya (Kenya)<br /> **Old output: &lt;code&gt;^Kenya/Kenya$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kenya/Kenya&lt;n&gt;&lt;prop&gt;$&lt;/code&gt;<br /> <br /> *Support for unique reversive middle voice morphology<br /> **Example word: rũthiũrũrũkĩirie (surround, roughly)<br /> **Old output: &lt;code&gt;^makĩinũkae/*makĩinũka$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^makĩinũka/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for ĩ to y alternation in certain cases (implemented through twol)<br /> **Example word: yarĩ (copula form)<br /> **Old output: &lt;code&gt;^yarĩ/*yarĩ$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yarĩ/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p2&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p13&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl7&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p10&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *An attempt was made to implement object agreement morphology on verbs. Such an attempt would necessitate adding twoc rules for each noun class, and doing so made twoc take far too long to compile. This attempt was therefore abandoned, but if a workaround were to be found for twoc, the twoc rules and lexc code are commented out in their respective files.<br /> <br /> ====Additional Constraint Grammar Rules====<br /> <br /> *If I am a verb, demonstrative, or associative inflected to agree with a certain number/noun class, select me if I am preceded by an adjective preceded by a noun of that number/noun class.<br /> **Example: maaĩ maingĩ me (many waters are)<br /> **Output:&lt;pre&gt;<br /> &quot;&lt;Maaĩ&gt;&quot;<br /> &quot;aĩ&quot; n cl5 pl<br /> &quot;&lt;maingĩ&gt;&quot;<br /> &quot;ingĩ&quot; adj cl5 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl9p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl14 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl11p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl15 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl12p6 pl SELECT:27<br /> &quot;&lt;me&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:33<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/noun class, select me if I am preceded by a demonstrative preceded by a noun of that number/person/noun class.<br /> **Example: andũ acio nĩmagakorwo (those people will find)<br /> **Output:&lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;andũ&gt;&quot;<br /> &quot;ndũ&quot; n cl1 pl<br /> &quot;&lt;acio&gt;&quot;<br /> &quot;pro&quot; prn dem ana cl1 pl<br /> &quot;&lt;nĩmagakorwo&gt;&quot;<br /> &quot;kor&quot; v pres pass p3 cl1 pl foc SELECT:40<br /> &quot;kor&quot; v remfut pass p3 cl1 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl12p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl12p6 pl foc SELECT:40<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/person/noun class, select me if I am preceded by a number preceded by an associative preceded by a noun of that noun class.<br /> **Example: nyamũ ya mũthia nĩ (the last animal is)<br /> **Output:<br /> &lt;pre&gt;<br /> &quot;&lt;nyamũ&gt;&quot;<br /> &quot;nyamũ&quot; n cl9p10 pl<br /> &quot;nyamũ&quot; n cl9p10 sg<br /> &quot;&lt;ya&gt;&quot;<br /> &quot;a&quot; assoc cl9p10 sg SELECT:27<br /> ; &quot;a&quot; assoc cl3 pl SELECT:27<br /> ; &quot;a&quot; assoc cl9p6 sg SELECT:27<br /> &quot;&lt;mũthia&gt;&quot;<br /> &quot;mũthia&quot; num<br /> &quot;&lt;nĩ&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl9p10 pl SELECT:46<br /> &quot;rĩ&quot; cop pres p3 cl9p10 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 sg SELECT:46<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a preposition followed by a verb, remove me.<br /> **Example: na ndĩarĩ na kĩndũ (and was of nothing)<br /> **Output:&lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> ; &quot;na&quot; pr REMOVE:68<br /> &quot;&lt;ndĩarĩ&gt;&quot;<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p6 sg<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p10 sg<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;kĩndũ&gt;&quot;<br /> &quot;ndũ&quot; n cl7 sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> <br /> <br /> ===kik → lin Implementation Details===<br /> This direction of the RBMT system was improved by adding items to the bilingual dictionary, adding additional morphology to the transducer, writing more constraint grammar rules, and writing additional transfer rules. Specific implementation details with examples of improved output are detailed in the sections that follow.<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for proper nouns<br /> **Example sentence: &quot;Nzambe&quot; (God)<br /> **Old output: &lt;code&gt;^Nzambe/*Nzambe$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nzambe/Nzambe&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for imperative mood<br /> **Example sentence: &quot;Kíma!&quot; (Run!)<br /> **Old output: &lt;code&gt;^Kíma/*Kíma$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kíma/kokíma&lt;v&gt;&lt;imp&gt;$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for intensifying suffixes (-a becomes -áká to denote certainty (infinitive, future) or insistence (imperative))<br /> **Example sentence 1: &quot;Azalákí kopésáká&quot; (He was really distributing...)<br /> **Old output: &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/*kopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/kopésa&lt;v&gt;&lt;inf&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 2: &quot;Akopésáká&quot; (He will really give...)<br /> **Old output: &lt;code&gt;^Akopésáká/*Akopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Akopésáká/kopésa&lt;v&gt;&lt;intns&gt;&lt;fut&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 3: &quot;Pésáká&quot; (Now do give...)<br /> **Old output: &lt;code&gt;^Pésáká/*Pésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Pésáká/kopésa&lt;v&gt;&lt;imp&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for more than 1 radical extension<br /> **Example sentence: &quot;Nabángisamákí na Pierre&quot; (I was scared by Pierre, or more literally, I was made, by Pierre, to fear something/someone)<br /> **Old output: &lt;code&gt;^Nabángisamákí/*Nabángisamákí$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nabángisamákí/kobánga&lt;v&gt;&lt;caus&gt;&lt;pass&gt;&lt;urp&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for numerals<br /> **Example sentence: &quot;zómi&quot; (ten)<br /> **Old output: &lt;code&gt;^zómi/*zómi$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^zómi/zómi&lt;num&gt;&lt;ord&gt;/zómi&lt;num&gt;&lt;card&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for quantifiers<br /> **Example sentence: &quot;mbelí nyónso&quot; (each knife)<br /> **Old output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/*nyónso$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/nyónso&lt;det&gt;&lt;qnt&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> ====Additional Constraint Grammar Rules====<br /> *If I am an adjective and a quantifier but I do not agree in number with the preceding noun, I must be a quantifier<br /> **Example sentence: &quot;bilóko moké&quot; (A few things, &#039;&#039;not&#039;&#039; small things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> &quot;moké&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> ; &quot;moké&quot; adj sg REMOVE:21<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a singular adjective, I cannot be plural<br /> **Example sentence: &quot;nzelá molaí&quot; (a long road)<br /> **Old output (below) <br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> ; &quot;nzelá&quot; n cl9 pl REMOVE:25<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a plural adjective, I cannot be singular<br /> **Example sentence: &quot;nzelá milaí&quot; (long roads)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 pl<br /> ; &quot;nzelá&quot; n cl9 sg REMOVE:28<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be an ordinal and a cardinal and am modifying a plural noun, I am not an ordinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;bilóko míbalé&quot; (two things, &#039;&#039;not&#039;&#039; the second thing)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num card<br /> ; &quot;míbalé&quot; num ord REMOVE:31<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be and ordinal and a cardinal and am modifying a a singular noun, I am not a cardinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;elóko mibalé&quot; (the second thing, &#039;&#039;not&#039;&#039; two things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> ; &quot;míbalé&quot; num card REMOVE:34<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be a demonstrative nominal and a relative pronoun and am proceeded by a noun, I cannot be a demonstrative nominal.<br /> **Example sentence: &quot;Mobali óyo amonaki ngáí&quot; (The man who saw me...)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> &quot;óyo&quot; prn dem sg<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> ; &quot;óyo&quot; prn dem sg REMOVE:37<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> *Enhance existing basic verb transfer rule to work with newly implemented imperative<br /> **Example sentence: &quot;ina&quot; (sing)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#koyemba&lt;/code&gt;<br /> **New output<br /> ***Tagger:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;imp&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;yemba&lt;/code&gt;<br /> <br /> *Map Kikuyu associative to Lingala connective by removing extra agreement tags found in Kikuyu<br /> **Example sentence: &quot;wa&quot;<br /> **Old output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#ya&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;ya&lt;/code&gt;<br /> <br /> *Basic adjective transfer: keep number markings if a variable adjective in Lingala, discard otherwise<br /> **Example sentence 1: anene (big)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#monéne&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;minéne&lt;/code&gt;<br /> **Example sentence 2: mata (sour)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#bololo&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;bololo&lt;/code&gt;<br /> <br /> ===lin Transducer Evaluation===<br /> *Precision (annotated corpus): 80.53%<br /> *Recall (annotated corpus): 55.64%<br /> *Coverage (large corpus): 62.10%<br /> *Number of words (large corpus): 275,546 (actual), 304,425 (tokenized)<br /> *Number of stems in transducer: 273<br /> <br /> ===kik Transducer Evaluation===<br /> *Precision (annotated corpus):<br /> *Recall (annotated corpus): <br /> *Coverage (large corpus):<br /> *Number of words (large corpus):<br /> *Number of stems in transducer:<br /> <br /> ===lin → kik MT Evaluation===<br /> *WER (longer corpus):<br /> *PER (longer corpus):<br /> *Percentage unknown words:<br /> *Percentage known words:<br /> *Trimmed coverage over longer corpus [lin-kik.automorf.bin applied to lin.longer.txt]: 89.29%<br /> *Trimmed coverage over large corpus [lin-kik.automorf.bin applied to lin.corpus.large.txt]: 48.90%<br /> *Number of stems in longer corpus: 117 (actual) 140 (tokenized, perhaps including punctuation)<br /> *Number of stems in large corpus: 275,546 (actual), 304,425 (tokenized)<br /> <br /> ===kik → lin MT Evaluation===<br /> *WER (longer corpus): 86.32%<br /> *PER (longer corpus): 75.21%<br /> *Percentage unknown words: 39.50% [Actually probably a bit higher since some words were marked with an &quot;@&quot; and not translated while the script only looks for words marked with a &quot;*&quot;]<br /> *Percentage known words: 60.60%<br /> *Trimmed coverage over longer corpus [kik-lin.automorf.bin applied to kik.corpus.larger.txt]:<br /> *Trimmed coverage over large corpus [kik-lin.automorf.bin applied to kik.longer.txt]:<br /> *Number of stems in longer corpus:<br /> *Number of stems in large corpus:<br /> <br /> <br /> [[Category:Sp17_TranslationPairs]]<br /> [[Category:Lingala]]<br /> [[Category:Kikuyu]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Lingala_and_Kikuyu&diff=4568 2017-04-17T21:06:48Z

<p>Jmundo1: </p> <hr /> <div>Resources for machine translation between [[Lingala]] and [[Kikuyu]]. Code is stored in a [https://github.swarthmore.edu/apowell1/ling073-kik-lin GitHub Repository].<br /> ==Developed Resources==<br /> ===Parallel Corpus===<br /> <br /> A [https://github.swarthmore.edu/apowell1/ling073-lin-kik-corpus parallel corpus] created using two translations of Genesis is available on Github.<br /> <br /> ===Contrastive Grammar===<br /> Contrastive Grammar is detailed on the wiki page [[Lingala and Kikuyu/Contrastive Grammar]].<br /> <br /> ===Lexical Selection===<br /> Lexical selection is detailed on the wiki page [[Lingala and Kikuyu/Lexical Selection]].<br /> <br /> ===Structural Transfer===<br /> Structural transfer is detailed on the wiki page [[Lingala and Kikuyu/Structural Transfer]]<br /> <br /> ==Installing Unidecode For MT Evaluation==<br /> Most available Lingala text does not mark tones; however, the Lingala transducer has been configured using spellrelax to provide analyses and surface generation with tone markings included as accents on vowels. As a result, generated translations will include tone markings not present in the reference translation and artificially inflate evaluation metrics such as WER and PER. Fortunately, we can take advantage of the fact that accented characters are not present in ASCII to replace accented characters with their unaccented version on the command line.<br /> <br /> [https://pypi.python.org/pypi/Unidecode Unidecode] is a Python package that implements a function to convert Unicode strings into their closest ASCII approximation. For simple accented characters, this is precisely the mapping we need to remove the accents for evaluation against a reference translation. Unidecode can also be used on the command line with our machine translation system as follows: &lt;code&gt;echo /path/to/kik-corpus | apertium -d . kik-lin | unidecode&lt;/code&gt;. To install unidecode, it is recommended to use &lt;code&gt;pip&lt;/code&gt;, as experience suggests that &lt;code&gt;apt-get&lt;/code&gt; will not run the necessary installation scripts to make unidecode accessible in Bash.<br /> <br /> On the Ling 073 virtual machines, Pip can be installed using &lt;code&gt;sudo apt-get install python-pip&lt;/code&gt;. After installing pip, unidecode can be installed using &lt;code&gt; pip install unidecode&lt;/code&gt;. <br /> <br /> ==Lexical Selection==<br /> This section is part of the [[Lexical transfer]] assignment. In addition to adding many one-to-one mappings to our .dix file, we have included a couple of one-to-many mappings for each direction of translation. See [[Lingala and Kikuyu/Lexical Selection]] for further details.<br /> <br /> From Lingala to Kikuyu, we consider four one-to-many mappings:<br /> *The Lingala verb kokíma can be translated in English as both &quot;to run&quot; and &quot;to run away.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, hanyũk and ũr.<br /> *The Lingala verb kofánda can be translated in English as both &quot;to sit&quot; and &quot;to live&quot; (as in reside). Kikuyu has two distinct verbs corresponding to these English translations, respectively, ikar and tũũr.<br /> *The Lingala verb kobánza can be translated in English as both &quot;to think&quot; and &quot;to remember.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, ĩcir and ririkan.<br /> *The Lingala noun mabele can be translated in English as both &quot;sand&quot; and &quot;milk.&quot; Kikuyu has two distinct nouns corresponding to these English translations, respectively, thanga and ria.<br /> <br /> From Kikuyu to Lingala, we consider three one-to-many-mappings:<br /> *The Kikuyu verb igu can be translated in English as both &quot;to obey&quot; and &quot;to hear.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, kotosa and koyoka.<br /> *The Kikuyu verb rut can be translated in English as both &quot;to learn&quot; and &quot;to remove.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, koyekola and kokukula.<br /> *the Kikuyu noun ria can be translated in English as both &quot;milk&quot; and &quot;pond.&quot; Lingala has two distinct nouns corresponding to these English translations, respectively, mabele and etimá.<br /> <br /> ==Initial Evaluation==<br /> ===lin → kik evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 87.50%<br /> **PER: 87.50%<br /> **Number of words: 9<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 99.06%<br /> **PER: 94.34%<br /> **Number of words: 107<br /> **Number of unknown words: 71<br /> **Proportion of stems translated correctly:33.64%<br /> <br /> ===kik → lin evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 80%<br /> **PER: 80%<br /> **Number of words: 7 (This should really be 8, but the command we are supposed to use to count this outputs 7. The proportion remains unchanged in our case, however.)<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 98.13%<br /> **PER: 90.65%<br /> **Number of words: 106<br /> **Number of unknown words: 79 (Note: Using the commands on the wiki gives a slightly different count that using the apertium-eval-translator tool. We use the count given by the command, not the tool)<br /> **Proportion of stems translated correctly: 25.47%<br /> <br /> ==Final Evaluation==<br /> ===lin → kik Implementation Details===<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for demonstrative pronouns/determiners (proximal, distal, and anaphoric)<br /> **Example word: &quot;rĩrĩa&quot; (Class 5 singular distal demonstrative)<br /> **Old output: &lt;code&gt;^rĩrĩa/*rĩrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^rĩrĩa/pro&lt;rel&gt;&lt;cl5&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dist&gt;&lt;cl5&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for possessive pronouns/determiners<br /> **Example word: &quot;gĩakwa&quot; (Possessive pronoun for Class 7 singular object with first-person singular owner)<br /> **Old output: &lt;code&gt;^gĩakwa/*gĩakwa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gĩakwa/pro&lt;prn&gt;&lt;poss&gt;&lt;cl7&gt;&lt;sg&gt;&lt;sp1&gt;&lt;ssg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for relative pronouns<br /> **Example word: ũrĩa (Class 1 singular relative pronoun, as well as other noun classes and other types of pronouns)<br /> **Old output: &lt;code&gt;^ũrĩa/*ũrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^ũrĩa/pro&lt;prn&gt;&lt;rel&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl14&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl14&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for personal pronouns<br /> **Example word: yo (Class 3 plural personal pronoun)<br /> **Old output: &lt;code&gt;^yo/*yo$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yo/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;sg&gt;$&lt;/code&gt;<br /> <br /> *Support for proper names (proper name tag introduced, and several names added)<br /> **Example word: Kenya (Kenya)<br /> **Old output: &lt;code&gt;^Kenya/Kenya$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kenya/Kenya&lt;n&gt;&lt;prop&gt;$&lt;/code&gt;<br /> <br /> *Support for unique reversive middle voice morphology<br /> **Example word: rũthiũrũrũkĩirie (surround, roughly)<br /> **Old output: &lt;code&gt;^makĩinũkae/*makĩinũka$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^makĩinũka/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for ĩ to y alternation in certain cases (implemented through twol)<br /> **Example word: yarĩ (copula form)<br /> **Old output: &lt;code&gt;^yarĩ/*yarĩ$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yarĩ/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p2&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p13&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl7&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p10&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *An attempt was made to implement object agreement morphology on verbs. Such an attempt would necessitate adding twoc rules for each noun class, and doing so made twoc take far too long to compile. This attempt was therefore abandoned, but if a workaround were to be found for twoc, the twoc rules and lexc code are commented out in their respective files.<br /> <br /> ====Additional Constraint Grammar Rules====<br /> <br /> *If I am a verb, demonstrative, or associative inflected to agree with a certain number/noun class, select me if I am preceded by an adjective preceded by a noun of that number/noun class.<br /> **Example: maaĩ maingĩ me (many waters are)<br /> **Output:&lt;pre&gt;<br /> &quot;&lt;Maaĩ&gt;&quot;<br /> &quot;aĩ&quot; n cl5 pl<br /> &quot;&lt;maingĩ&gt;&quot;<br /> &quot;ingĩ&quot; adj cl5 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl9p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl14 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl11p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl15 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl12p6 pl SELECT:27<br /> &quot;&lt;me&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:33<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/noun class, select me if I am preceded by a demonstrative preceded by a noun of that number/person/noun class.<br /> **Example: andũ acio nĩmagakorwo (those people will find)<br /> **Output:&lt;pre&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;andũ&gt;&quot;<br /> &quot;ndũ&quot; n cl1 pl<br /> &quot;&lt;acio&gt;&quot;<br /> &quot;pro&quot; prn dem ana cl1 pl<br /> &quot;&lt;nĩmagakorwo&gt;&quot;<br /> &quot;kor&quot; v pres pass p3 cl1 pl foc SELECT:40<br /> &quot;kor&quot; v remfut pass p3 cl1 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl12p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl12p6 pl foc SELECT:40<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/person/noun class, select me if I am preceded by a number preceded by an associative preceded by a noun of that noun class.<br /> **Example: nyamũ ya mũthia nĩ (the last animal is)<br /> **Output:&lt;pre&gt;<br /> &quot;&lt;nyamũ&gt;&quot;<br /> &quot;nyamũ&quot; n cl9p10 pl<br /> &quot;nyamũ&quot; n cl9p10 sg<br /> &quot;&lt;ya&gt;&quot;<br /> &quot;a&quot; assoc cl9p10 sg SELECT:27<br /> ; &quot;a&quot; assoc cl3 pl SELECT:27<br /> ; &quot;a&quot; assoc cl9p6 sg SELECT:27<br /> &quot;&lt;mũthia&gt;&quot;<br /> &quot;mũthia&quot; num<br /> &quot;&lt;nĩ&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl9p10 pl SELECT:46<br /> &quot;rĩ&quot; cop pres p3 cl9p10 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 sg SELECT:46<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a preposition followed by a verb, remove me.<br /> **Example: na ndĩarĩ na kĩndũ (and was of nothing)<br /> **Output:&lt;code&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> ; &quot;na&quot; pr REMOVE:68<br /> &quot;&lt;ndĩarĩ&gt;&quot;<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p6 sg<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p10 sg<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;kĩndũ&gt;&quot;<br /> &quot;ndũ&quot; n cl7 sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/code&gt;<br /> <br /> ====Additional Transfer Rules====<br /> <br /> <br /> ===kik → lin Implementation Details===<br /> This direction of the RBMT system was improved by adding items to the bilingual dictionary, adding additional morphology to the transducer, writing more constraint grammar rules, and writing additional transfer rules. Specific implementation details with examples of improved output are detailed in the sections that follow.<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for proper nouns<br /> **Example sentence: &quot;Nzambe&quot; (God)<br /> **Old output: &lt;code&gt;^Nzambe/*Nzambe$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nzambe/Nzambe&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for imperative mood<br /> **Example sentence: &quot;Kíma!&quot; (Run!)<br /> **Old output: &lt;code&gt;^Kíma/*Kíma$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kíma/kokíma&lt;v&gt;&lt;imp&gt;$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for intensifying suffixes (-a becomes -áká to denote certainty (infinitive, future) or insistence (imperative))<br /> **Example sentence 1: &quot;Azalákí kopésáká&quot; (He was really distributing...)<br /> **Old output: &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/*kopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/kopésa&lt;v&gt;&lt;inf&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 2: &quot;Akopésáká&quot; (He will really give...)<br /> **Old output: &lt;code&gt;^Akopésáká/*Akopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Akopésáká/kopésa&lt;v&gt;&lt;intns&gt;&lt;fut&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 3: &quot;Pésáká&quot; (Now do give...)<br /> **Old output: &lt;code&gt;^Pésáká/*Pésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Pésáká/kopésa&lt;v&gt;&lt;imp&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for more than 1 radical extension<br /> **Example sentence: &quot;Nabángisamákí na Pierre&quot; (I was scared by Pierre, or more literally, I was made, by Pierre, to fear something/someone)<br /> **Old output: &lt;code&gt;^Nabángisamákí/*Nabángisamákí$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nabángisamákí/kobánga&lt;v&gt;&lt;caus&gt;&lt;pass&gt;&lt;urp&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for numerals<br /> **Example sentence: &quot;zómi&quot; (ten)<br /> **Old output: &lt;code&gt;^zómi/*zómi$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^zómi/zómi&lt;num&gt;&lt;ord&gt;/zómi&lt;num&gt;&lt;card&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for quantifiers<br /> **Example sentence: &quot;mbelí nyónso&quot; (each knife)<br /> **Old output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/*nyónso$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/nyónso&lt;det&gt;&lt;qnt&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> ====Additional Constraint Grammar Rules====<br /> *If I am an adjective and a quantifier but I do not agree in number with the preceding noun, I must be a quantifier<br /> **Example sentence: &quot;bilóko moké&quot; (A few things, &#039;&#039;not&#039;&#039; small things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> &quot;moké&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> ; &quot;moké&quot; adj sg REMOVE:21<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a singular adjective, I cannot be plural<br /> **Example sentence: &quot;nzelá molaí&quot; (a long road)<br /> **Old output (below) <br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> ; &quot;nzelá&quot; n cl9 pl REMOVE:25<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a plural adjective, I cannot be singular<br /> **Example sentence: &quot;nzelá milaí&quot; (long roads)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 pl<br /> ; &quot;nzelá&quot; n cl9 sg REMOVE:28<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be an ordinal and a cardinal and am modifying a plural noun, I am not an ordinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;bilóko míbalé&quot; (two things, &#039;&#039;not&#039;&#039; the second thing)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num card<br /> ; &quot;míbalé&quot; num ord REMOVE:31<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be and ordinal and a cardinal and am modifying a a singular noun, I am not a cardinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;elóko mibalé&quot; (the second thing, &#039;&#039;not&#039;&#039; two things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> ; &quot;míbalé&quot; num card REMOVE:34<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be a demonstrative nominal and a relative pronoun and am proceeded by a noun, I cannot be a demonstrative nominal.<br /> **Example sentence: &quot;Mobali óyo amonaki ngáí&quot; (The man who saw me...)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> &quot;óyo&quot; prn dem sg<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> ; &quot;óyo&quot; prn dem sg REMOVE:37<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> *Enhance existing basic verb transfer rule to work with newly implemented imperative<br /> **Example sentence: &quot;ina&quot; (sing)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#koyemba&lt;/code&gt;<br /> **New output<br /> ***Tagger:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;imp&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;yemba&lt;/code&gt;<br /> <br /> *Map Kikuyu associative to Lingala connective by removing extra agreement tags found in Kikuyu<br /> **Example sentence: &quot;wa&quot;<br /> **Old output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#ya&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;ya&lt;/code&gt;<br /> <br /> *Basic adjective transfer: keep number markings if a variable adjective in Lingala, discard otherwise<br /> **Example sentence 1: anene (big)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#monéne&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;minéne&lt;/code&gt;<br /> **Example sentence 2: mata (sour)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#bololo&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;bololo&lt;/code&gt;<br /> <br /> ===lin Transducer Evaluation===<br /> *Precision (annotated corpus): 80.53%<br /> *Recall (annotated corpus): 55.64%<br /> *Coverage (large corpus): 62.10%<br /> *Number of words (large corpus): 275,546 (actual), 304,425 (tokenized)<br /> *Number of stems in transducer: 273<br /> <br /> ===kik Transducer Evaluation===<br /> *Precision (annotated corpus):<br /> *Recall (annotated corpus): <br /> *Coverage (large corpus):<br /> *Number of words (large corpus):<br /> *Number of stems in transducer:<br /> <br /> ===lin → kik MT Evaluation===<br /> *WER (longer corpus):<br /> *PER (longer corpus):<br /> *Percentage unknown words:<br /> *Percentage known words:<br /> *Trimmed coverage over longer corpus [lin-kik.automorf.bin applied to lin.longer.txt]: 89.29%<br /> *Trimmed coverage over large corpus [lin-kik.automorf.bin applied to lin.corpus.large.txt]: 48.90%<br /> *Number of stems in longer corpus: 117 (actual) 140 (tokenized, perhaps including punctuation)<br /> *Number of stems in large corpus: 275,546 (actual), 304,425 (tokenized)<br /> <br /> ===kik → lin MT Evaluation===<br /> *WER (longer corpus): 86.32%<br /> *PER (longer corpus): 75.21%<br /> *Percentage unknown words: 39.50% [Actually probably a bit higher since some words were marked with an &quot;@&quot; and not translated while the script only looks for words marked with a &quot;*&quot;]<br /> *Percentage known words: 60.60%<br /> *Trimmed coverage over longer corpus [kik-lin.automorf.bin applied to kik.corpus.larger.txt]:<br /> *Trimmed coverage over large corpus [kik-lin.automorf.bin applied to kik.longer.txt]:<br /> *Number of stems in longer corpus:<br /> *Number of stems in large corpus:<br /> <br /> <br /> [[Category:Sp17_TranslationPairs]]<br /> [[Category:Lingala]]<br /> [[Category:Kikuyu]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Lingala_and_Kikuyu&diff=4567 2017-04-17T21:05:32Z

<p>Jmundo1: </p> <hr /> <div>Resources for machine translation between [[Lingala]] and [[Kikuyu]]. Code is stored in a [https://github.swarthmore.edu/apowell1/ling073-kik-lin GitHub Repository].<br /> ==Developed Resources==<br /> ===Parallel Corpus===<br /> <br /> A [https://github.swarthmore.edu/apowell1/ling073-lin-kik-corpus parallel corpus] created using two translations of Genesis is available on Github.<br /> <br /> ===Contrastive Grammar===<br /> Contrastive Grammar is detailed on the wiki page [[Lingala and Kikuyu/Contrastive Grammar]].<br /> <br /> ===Lexical Selection===<br /> Lexical selection is detailed on the wiki page [[Lingala and Kikuyu/Lexical Selection]].<br /> <br /> ===Structural Transfer===<br /> Structural transfer is detailed on the wiki page [[Lingala and Kikuyu/Structural Transfer]]<br /> <br /> ==Installing Unidecode For MT Evaluation==<br /> Most available Lingala text does not mark tones; however, the Lingala transducer has been configured using spellrelax to provide analyses and surface generation with tone markings included as accents on vowels. As a result, generated translations will include tone markings not present in the reference translation and artificially inflate evaluation metrics such as WER and PER. Fortunately, we can take advantage of the fact that accented characters are not present in ASCII to replace accented characters with their unaccented version on the command line.<br /> <br /> [https://pypi.python.org/pypi/Unidecode Unidecode] is a Python package that implements a function to convert Unicode strings into their closest ASCII approximation. For simple accented characters, this is precisely the mapping we need to remove the accents for evaluation against a reference translation. Unidecode can also be used on the command line with our machine translation system as follows: &lt;code&gt;echo /path/to/kik-corpus | apertium -d . kik-lin | unidecode&lt;/code&gt;. To install unidecode, it is recommended to use &lt;code&gt;pip&lt;/code&gt;, as experience suggests that &lt;code&gt;apt-get&lt;/code&gt; will not run the necessary installation scripts to make unidecode accessible in Bash.<br /> <br /> On the Ling 073 virtual machines, Pip can be installed using &lt;code&gt;sudo apt-get install python-pip&lt;/code&gt;. After installing pip, unidecode can be installed using &lt;code&gt; pip install unidecode&lt;/code&gt;. <br /> <br /> ==Lexical Selection==<br /> This section is part of the [[Lexical transfer]] assignment. In addition to adding many one-to-one mappings to our .dix file, we have included a couple of one-to-many mappings for each direction of translation. See [[Lingala and Kikuyu/Lexical Selection]] for further details.<br /> <br /> From Lingala to Kikuyu, we consider four one-to-many mappings:<br /> *The Lingala verb kokíma can be translated in English as both &quot;to run&quot; and &quot;to run away.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, hanyũk and ũr.<br /> *The Lingala verb kofánda can be translated in English as both &quot;to sit&quot; and &quot;to live&quot; (as in reside). Kikuyu has two distinct verbs corresponding to these English translations, respectively, ikar and tũũr.<br /> *The Lingala verb kobánza can be translated in English as both &quot;to think&quot; and &quot;to remember.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, ĩcir and ririkan.<br /> *The Lingala noun mabele can be translated in English as both &quot;sand&quot; and &quot;milk.&quot; Kikuyu has two distinct nouns corresponding to these English translations, respectively, thanga and ria.<br /> <br /> From Kikuyu to Lingala, we consider three one-to-many-mappings:<br /> *The Kikuyu verb igu can be translated in English as both &quot;to obey&quot; and &quot;to hear.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, kotosa and koyoka.<br /> *The Kikuyu verb rut can be translated in English as both &quot;to learn&quot; and &quot;to remove.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, koyekola and kokukula.<br /> *the Kikuyu noun ria can be translated in English as both &quot;milk&quot; and &quot;pond.&quot; Lingala has two distinct nouns corresponding to these English translations, respectively, mabele and etimá.<br /> <br /> ==Initial Evaluation==<br /> ===lin → kik evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 87.50%<br /> **PER: 87.50%<br /> **Number of words: 9<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 99.06%<br /> **PER: 94.34%<br /> **Number of words: 107<br /> **Number of unknown words: 71<br /> **Proportion of stems translated correctly:33.64%<br /> <br /> ===kik → lin evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 80%<br /> **PER: 80%<br /> **Number of words: 7 (This should really be 8, but the command we are supposed to use to count this outputs 7. The proportion remains unchanged in our case, however.)<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 98.13%<br /> **PER: 90.65%<br /> **Number of words: 106<br /> **Number of unknown words: 79 (Note: Using the commands on the wiki gives a slightly different count that using the apertium-eval-translator tool. We use the count given by the command, not the tool)<br /> **Proportion of stems translated correctly: 25.47%<br /> <br /> ==Final Evaluation==<br /> ===lin → kik Implementation Details===<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for demonstrative pronouns/determiners (proximal, distal, and anaphoric)<br /> **Example word: &quot;rĩrĩa&quot; (Class 5 singular distal demonstrative)<br /> **Old output: &lt;code&gt;^rĩrĩa/*rĩrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^rĩrĩa/pro&lt;rel&gt;&lt;cl5&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dist&gt;&lt;cl5&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for possessive pronouns/determiners<br /> **Example word: &quot;gĩakwa&quot; (Possessive pronoun for Class 7 singular object with first-person singular owner)<br /> **Old output: &lt;code&gt;^gĩakwa/*gĩakwa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gĩakwa/pro&lt;prn&gt;&lt;poss&gt;&lt;cl7&gt;&lt;sg&gt;&lt;sp1&gt;&lt;ssg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for relative pronouns<br /> **Example word: ũrĩa (Class 1 singular relative pronoun, as well as other noun classes and other types of pronouns)<br /> **Old output: &lt;code&gt;^ũrĩa/*ũrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^ũrĩa/pro&lt;prn&gt;&lt;rel&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;rel&gt;&lt;cl14&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl1&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl3&gt;&lt;sg&gt;/pro&lt;prn&gt;&lt;dem&gt;&lt;dist&gt;&lt;cl14&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for personal pronouns<br /> **Example word: yo (Class 3 plural personal pronoun)<br /> **Old output: &lt;code&gt;^yo/*yo$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yo/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/pro&lt;prn&gt;&lt;pers&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;sg&gt;$&lt;/code&gt;<br /> <br /> *Support for proper names (proper name tag introduced, and several names added)<br /> **Example word: Kenya (Kenya)<br /> **Old output: &lt;code&gt;^Kenya/Kenya$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kenya/Kenya&lt;n&gt;&lt;prop&gt;$&lt;/code&gt;<br /> <br /> *Support for unique reversive middle voice morphology<br /> **Example word: rũthiũrũrũkĩirie (surround, roughly)<br /> **Old output: &lt;code&gt;^makĩinũkae/*makĩinũka$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^makĩinũka/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/in&lt;seq&gt;&lt;mid&gt;&lt;rev&gt;&lt;refl&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for ĩ to y alternation in certain cases (implemented through twol)<br /> **Example word: yarĩ (copula form)<br /> **Old output: &lt;code&gt;^yarĩ/*yarĩ$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^yarĩ/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p2&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl1&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl3&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl14&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl5&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl15&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p13&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl7&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl12p6&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl9p10&gt;&lt;pl&gt;/rĩ&lt;cop&gt;&lt;rempast&gt;&lt;p3&gt;&lt;cl11p10&gt;&lt;pl&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *An attempt was made to implement object agreement morphology on verbs. Such an attempt would necessitate adding twoc rules for each noun class, and doing so made twoc take far too long to compile. This attempt was therefore abandoned, but if a workaround were to be found for twoc, the twoc rules and lexc code are commented out in their respective files.<br /> <br /> ====Additional Constraint Grammar Rules====<br /> <br /> *If I am a verb, demonstrative, or associative inflected to agree with a certain number/noun class, select me if I am preceded by an adjective preceded by a noun of that number/noun class.<br /> **Example: maaĩ maingĩ me (many waters are)<br /> **Output:&lt;code&gt;<br /> &quot;&lt;Maaĩ&gt;&quot;<br /> &quot;aĩ&quot; n cl5 pl<br /> &quot;&lt;maingĩ&gt;&quot;<br /> &quot;ingĩ&quot; adj cl5 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl9p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl14 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl11p6 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl15 pl SELECT:27<br /> ; &quot;ingĩ&quot; adj cl12p6 pl SELECT:27<br /> &quot;&lt;me&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:33<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:33<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/code&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/noun class, select me if I am preceded by a demonstrative preceded by a noun of that number/person/noun class.<br /> **Example: andũ acio nĩmagakorwo (those people will find)<br /> **Output:&lt;code&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;andũ&gt;&quot;<br /> &quot;ndũ&quot; n cl1 pl<br /> &quot;&lt;acio&gt;&quot;<br /> &quot;pro&quot; prn dem ana cl1 pl<br /> &quot;&lt;nĩmagakorwo&gt;&quot;<br /> &quot;kor&quot; v pres pass p3 cl1 pl foc SELECT:40<br /> &quot;kor&quot; v remfut pass p3 cl1 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v pres pass p3 cl12p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl9p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl14 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl5 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl11p6 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl15 pl foc SELECT:40<br /> ; &quot;kor&quot; v remfut pass p3 cl12p6 pl foc SELECT:40<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/code&gt;<br /> <br /> *If I am a verb inflected to agree with a certain number/person/noun class, select me if I am preceded by a number preceded by an associative preceded by a noun of that noun class.<br /> **Example: nyamũ ya mũthia nĩ (the last animal is)<br /> **Output:&lt;code&gt;<br /> &quot;&lt;nyamũ&gt;&quot;<br /> &quot;nyamũ&quot; n cl9p10 pl<br /> &quot;nyamũ&quot; n cl9p10 sg<br /> &quot;&lt;ya&gt;&quot;<br /> &quot;a&quot; assoc cl9p10 sg SELECT:27<br /> ; &quot;a&quot; assoc cl3 pl SELECT:27<br /> ; &quot;a&quot; assoc cl9p6 sg SELECT:27<br /> &quot;&lt;mũthia&gt;&quot;<br /> &quot;mũthia&quot; num<br /> &quot;&lt;nĩ&gt;&quot;<br /> &quot;rĩ&quot; cop pres p3 cl9p10 pl SELECT:46<br /> &quot;rĩ&quot; cop pres p3 cl9p10 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl1 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl3 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl9p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl14 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl5 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl15 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p13 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl7 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl12p6 sg SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 pl SELECT:46<br /> ; &quot;rĩ&quot; cop pres p3 cl11p10 sg SELECT:46<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/code&gt;<br /> <br /> *If I am a preposition followed by a verb, remove me.<br /> **Example: na ndĩarĩ na kĩndũ (and was of nothing)<br /> **Output:&lt;code&gt;<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> ; &quot;na&quot; pr REMOVE:68<br /> &quot;&lt;ndĩarĩ&gt;&quot;<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p6 sg<br /> &quot;rĩ&quot; cop rempast neg p3 cl9p10 sg<br /> &quot;&lt;na&gt;&quot;<br /> &quot;na&quot; cnjcoo<br /> &quot;na&quot; pr<br /> &quot;&lt;kĩndũ&gt;&quot;<br /> &quot;ndũ&quot; n cl7 sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/code&gt;<br /> <br /> ====Additional Transfer Rules====<br /> <br /> <br /> ===kik → lin Implementation Details===<br /> This direction of the RBMT system was improved by adding items to the bilingual dictionary, adding additional morphology to the transducer, writing more constraint grammar rules, and writing additional transfer rules. Specific implementation details with examples of improved output are detailed in the sections that follow.<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for proper nouns<br /> **Example sentence: &quot;Nzambe&quot; (God)<br /> **Old output: &lt;code&gt;^Nzambe/*Nzambe$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nzambe/Nzambe&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for imperative mood<br /> **Example sentence: &quot;Kíma!&quot; (Run!)<br /> **Old output: &lt;code&gt;^Kíma/*Kíma$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kíma/kokíma&lt;v&gt;&lt;imp&gt;$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for intensifying suffixes (-a becomes -áká to denote certainty (infinitive, future) or insistence (imperative))<br /> **Example sentence 1: &quot;Azalákí kopésáká&quot; (He was really distributing...)<br /> **Old output: &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/*kopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/kopésa&lt;v&gt;&lt;inf&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 2: &quot;Akopésáká&quot; (He will really give...)<br /> **Old output: &lt;code&gt;^Akopésáká/*Akopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Akopésáká/kopésa&lt;v&gt;&lt;intns&gt;&lt;fut&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 3: &quot;Pésáká&quot; (Now do give...)<br /> **Old output: &lt;code&gt;^Pésáká/*Pésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Pésáká/kopésa&lt;v&gt;&lt;imp&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for more than 1 radical extension<br /> **Example sentence: &quot;Nabángisamákí na Pierre&quot; (I was scared by Pierre, or more literally, I was made, by Pierre, to fear something/someone)<br /> **Old output: &lt;code&gt;^Nabángisamákí/*Nabángisamákí$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nabángisamákí/kobánga&lt;v&gt;&lt;caus&gt;&lt;pass&gt;&lt;urp&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for numerals<br /> **Example sentence: &quot;zómi&quot; (ten)<br /> **Old output: &lt;code&gt;^zómi/*zómi$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^zómi/zómi&lt;num&gt;&lt;ord&gt;/zómi&lt;num&gt;&lt;card&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for quantifiers<br /> **Example sentence: &quot;mbelí nyónso&quot; (each knife)<br /> **Old output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/*nyónso$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/nyónso&lt;det&gt;&lt;qnt&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> ====Additional Constraint Grammar Rules====<br /> *If I am an adjective and a quantifier but I do not agree in number with the preceding noun, I must be a quantifier<br /> **Example sentence: &quot;bilóko moké&quot; (A few things, &#039;&#039;not&#039;&#039; small things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> &quot;moké&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> ; &quot;moké&quot; adj sg REMOVE:21<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a singular adjective, I cannot be plural<br /> **Example sentence: &quot;nzelá molaí&quot; (a long road)<br /> **Old output (below) <br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> ; &quot;nzelá&quot; n cl9 pl REMOVE:25<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a plural adjective, I cannot be singular<br /> **Example sentence: &quot;nzelá milaí&quot; (long roads)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 pl<br /> ; &quot;nzelá&quot; n cl9 sg REMOVE:28<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be an ordinal and a cardinal and am modifying a plural noun, I am not an ordinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;bilóko míbalé&quot; (two things, &#039;&#039;not&#039;&#039; the second thing)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num card<br /> ; &quot;míbalé&quot; num ord REMOVE:31<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be and ordinal and a cardinal and am modifying a a singular noun, I am not a cardinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;elóko mibalé&quot; (the second thing, &#039;&#039;not&#039;&#039; two things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> ; &quot;míbalé&quot; num card REMOVE:34<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be a demonstrative nominal and a relative pronoun and am proceeded by a noun, I cannot be a demonstrative nominal.<br /> **Example sentence: &quot;Mobali óyo amonaki ngáí&quot; (The man who saw me...)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> &quot;óyo&quot; prn dem sg<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> ; &quot;óyo&quot; prn dem sg REMOVE:37<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> *Enhance existing basic verb transfer rule to work with newly implemented imperative<br /> **Example sentence: &quot;ina&quot; (sing)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#koyemba&lt;/code&gt;<br /> **New output<br /> ***Tagger:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;imp&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;yemba&lt;/code&gt;<br /> <br /> *Map Kikuyu associative to Lingala connective by removing extra agreement tags found in Kikuyu<br /> **Example sentence: &quot;wa&quot;<br /> **Old output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#ya&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;ya&lt;/code&gt;<br /> <br /> *Basic adjective transfer: keep number markings if a variable adjective in Lingala, discard otherwise<br /> **Example sentence 1: anene (big)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#monéne&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;minéne&lt;/code&gt;<br /> **Example sentence 2: mata (sour)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#bololo&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;bololo&lt;/code&gt;<br /> <br /> ===lin Transducer Evaluation===<br /> *Precision (annotated corpus): 80.53%<br /> *Recall (annotated corpus): 55.64%<br /> *Coverage (large corpus): 62.10%<br /> *Number of words (large corpus): 275,546 (actual), 304,425 (tokenized)<br /> *Number of stems in transducer: 273<br /> <br /> ===kik Transducer Evaluation===<br /> *Precision (annotated corpus):<br /> *Recall (annotated corpus): <br /> *Coverage (large corpus):<br /> *Number of words (large corpus):<br /> *Number of stems in transducer:<br /> <br /> ===lin → kik MT Evaluation===<br /> *WER (longer corpus):<br /> *PER (longer corpus):<br /> *Percentage unknown words:<br /> *Percentage known words:<br /> *Trimmed coverage over longer corpus [lin-kik.automorf.bin applied to lin.longer.txt]: 89.29%<br /> *Trimmed coverage over large corpus [lin-kik.automorf.bin applied to lin.corpus.large.txt]: 48.90%<br /> *Number of stems in longer corpus: 117 (actual) 140 (tokenized, perhaps including punctuation)<br /> *Number of stems in large corpus: 275,546 (actual), 304,425 (tokenized)<br /> <br /> ===kik → lin MT Evaluation===<br /> *WER (longer corpus): 86.32%<br /> *PER (longer corpus): 75.21%<br /> *Percentage unknown words: 39.50% [Actually probably a bit higher since some words were marked with an &quot;@&quot; and not translated while the script only looks for words marked with a &quot;*&quot;]<br /> *Percentage known words: 60.60%<br /> *Trimmed coverage over longer corpus [kik-lin.automorf.bin applied to kik.corpus.larger.txt]:<br /> *Trimmed coverage over large corpus [kik-lin.automorf.bin applied to kik.longer.txt]:<br /> *Number of stems in longer corpus:<br /> *Number of stems in large corpus:<br /> <br /> <br /> [[Category:Sp17_TranslationPairs]]<br /> [[Category:Lingala]]<br /> [[Category:Kikuyu]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Lingala_and_Kikuyu&diff=4566 2017-04-17T19:59:44Z

<p>Jmundo1: </p> <hr /> <div>Resources for machine translation between [[Lingala]] and [[Kikuyu]]. Code is stored in a [https://github.swarthmore.edu/apowell1/ling073-kik-lin GitHub Repository].<br /> ==Developed Resources==<br /> ===Parallel Corpus===<br /> <br /> A [https://github.swarthmore.edu/apowell1/ling073-lin-kik-corpus parallel corpus] created using two translations of Genesis is available on Github.<br /> <br /> ===Contrastive Grammar===<br /> Contrastive Grammar is detailed on the wiki page [[Lingala and Kikuyu/Contrastive Grammar]].<br /> <br /> ===Lexical Selection===<br /> Lexical selection is detailed on the wiki page [[Lingala and Kikuyu/Lexical Selection]].<br /> <br /> ===Structural Transfer===<br /> Structural transfer is detailed on the wiki page [[Lingala and Kikuyu/Structural Transfer]]<br /> <br /> ==Installing Unidecode For MT Evaluation==<br /> Most available Lingala text does not mark tones; however, the Lingala transducer has been configured using spellrelax to provide analyses and surface generation with tone markings included as accents on vowels. As a result, generated translations will include tone markings not present in the reference translation and artificially inflate evaluation metrics such as WER and PER. Fortunately, we can take advantage of the fact that accented characters are not present in ASCII to replace accented characters with their unaccented version on the command line.<br /> <br /> [https://pypi.python.org/pypi/Unidecode Unidecode] is a Python package that implements a function to convert Unicode strings into their closest ASCII approximation. For simple accented characters, this is precisely the mapping we need to remove the accents for evaluation against a reference translation. Unidecode can also be used on the command line with our machine translation system as follows: &lt;code&gt;echo /path/to/kik-corpus | apertium -d . kik-lin | unidecode&lt;/code&gt;. To install unidecode, it is recommended to use &lt;code&gt;pip&lt;/code&gt;, as experience suggests that &lt;code&gt;apt-get&lt;/code&gt; will not run the necessary installation scripts to make unidecode accessible in Bash.<br /> <br /> On the Ling 073 virtual machines, Pip can be installed using &lt;code&gt;sudo apt-get install python-pip&lt;/code&gt;. After installing pip, unidecode can be installed using &lt;code&gt; pip install unidecode&lt;/code&gt;. <br /> <br /> ==Lexical Selection==<br /> This section is part of the [[Lexical transfer]] assignment. In addition to adding many one-to-one mappings to our .dix file, we have included a couple of one-to-many mappings for each direction of translation. See [[Lingala and Kikuyu/Lexical Selection]] for further details.<br /> <br /> From Lingala to Kikuyu, we consider four one-to-many mappings:<br /> *The Lingala verb kokíma can be translated in English as both &quot;to run&quot; and &quot;to run away.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, hanyũk and ũr.<br /> *The Lingala verb kofánda can be translated in English as both &quot;to sit&quot; and &quot;to live&quot; (as in reside). Kikuyu has two distinct verbs corresponding to these English translations, respectively, ikar and tũũr.<br /> *The Lingala verb kobánza can be translated in English as both &quot;to think&quot; and &quot;to remember.&quot; Kikuyu has two distinct verbs corresponding to these English translations, respectively, ĩcir and ririkan.<br /> *The Lingala noun mabele can be translated in English as both &quot;sand&quot; and &quot;milk.&quot; Kikuyu has two distinct nouns corresponding to these English translations, respectively, thanga and ria.<br /> <br /> From Kikuyu to Lingala, we consider three one-to-many-mappings:<br /> *The Kikuyu verb igu can be translated in English as both &quot;to obey&quot; and &quot;to hear.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, kotosa and koyoka.<br /> *The Kikuyu verb rut can be translated in English as both &quot;to learn&quot; and &quot;to remove.&quot; Lingala has two distinct verbs corresponding to these English translations, respectively, koyekola and kokukula.<br /> *the Kikuyu noun ria can be translated in English as both &quot;milk&quot; and &quot;pond.&quot; Lingala has two distinct nouns corresponding to these English translations, respectively, mabele and etimá.<br /> <br /> ==Initial Evaluation==<br /> ===lin → kik evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 87.50%<br /> **PER: 87.50%<br /> **Number of words: 9<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 99.06%<br /> **PER: 94.34%<br /> **Number of words: 107<br /> **Number of unknown words: 71<br /> **Proportion of stems translated correctly:33.64%<br /> <br /> ===kik → lin evaluation===<br /> *Evaluation on mini corpus:<br /> **WER: 80%<br /> **PER: 80%<br /> **Number of words: 7 (This should really be 8, but the command we are supposed to use to count this outputs 7. The proportion remains unchanged in our case, however.)<br /> **Number of unknown words: 0<br /> **Proportion of stems translated correctly: 100%<br /> <br /> *Evaluation of corpus:<br /> **WER: 98.13%<br /> **PER: 90.65%<br /> **Number of words: 106<br /> **Number of unknown words: 79 (Note: Using the commands on the wiki gives a slightly different count that using the apertium-eval-translator tool. We use the count given by the command, not the tool)<br /> **Proportion of stems translated correctly: 25.47%<br /> <br /> ==Final Evaluation==<br /> ===lin → kik Implementation Details===<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for demonstrative pronouns/determiners (proximal, distal, and anaphoric)<br /> **Example word: &quot;rĩrĩa&quot; (Class 5 singular distal demonstrative)<br /> **Old output: &lt;code&gt;^rĩrĩa/*rĩrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^rĩrĩa/pro&lt;prn&gt;&lt;dist&gt;&lt;cl5&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for possessive pronouns/determiners<br /> **Example word: &quot;gĩakwa&quot; (Possessive pronoun for Class 7 singular object with first-person singular owner)<br /> **Old output: &lt;code&gt;^gĩakwa/*gĩakwa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^gĩakwa/pro&lt;prn&gt;&lt;poss&gt;&lt;cl7&gt;&lt;sg&gt;&lt;sp1&gt;&lt;ssg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for relative pronouns<br /> **Example word: ũrĩa (Class 1 singular relative pronoun)<br /> **Old output: &lt;code&gt;^ũrĩa/*ũrĩa$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^ũrĩa/pro&lt;prn&gt;&lt;rel&gt;&lt;cl1&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for unique reversive middle voice morphology<br /> **Example word: rũthiũrũrũkĩirie (surround, roughly)<br /> **Old output: &lt;code&gt;^rũthiũrũrũkĩirie/*rũthiũrũrũkĩirie$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^rũthiũrũrũkĩirie/thi&lt;v&gt;&lt;currpast&gt;&lt;intens&gt;&lt;revers&gt;&lt;mid&gt;&lt;app&gt;&lt;compl&gt;&lt;trans&gt;&lt;p3&gt;&lt;cl11p10&gt;&lt;sg&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *An attempt was made to implement object agreement morphology on verbs. Such an attempt would necessitate adding twoc rules for each noun class, and doing so made twoc take far too long to compile. This attempt was therefore abandoned, but if a workaround were to be found for twoc, the twoc rules and lexc code are commented out in their respective files.<br /> <br /> ===kik → lin Implementation Details===<br /> This direction of the RBMT system was improved by adding items to the bilingual dictionary, adding additional morphology to the transducer, writing more constraint grammar rules, and writing additional transfer rules. Specific implementation details with examples of improved output are detailed in the sections that follow.<br /> <br /> ====Bilingual Dictionary Expansion====<br /> Difference can be seen by looking at the Github commit history.<br /> <br /> ====Morphology Expansion====<br /> *Support for proper nouns<br /> **Example sentence: &quot;Nzambe&quot; (God)<br /> **Old output: &lt;code&gt;^Nzambe/*Nzambe$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nzambe/Nzambe&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for imperative mood<br /> **Example sentence: &quot;Kíma!&quot; (Run!)<br /> **Old output: &lt;code&gt;^Kíma/*Kíma$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Kíma/kokíma&lt;v&gt;&lt;imp&gt;$^!/!&lt;sent&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for intensifying suffixes (-a becomes -áká to denote certainty (infinitive, future) or insistence (imperative))<br /> **Example sentence 1: &quot;Azalákí kopésáká&quot; (He was really distributing...)<br /> **Old output: &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/*kopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Azalákí/kozala&lt;v&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;/kozala&lt;vaux&gt;&lt;urp&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$ ^kopésáká/kopésa&lt;v&gt;&lt;inf&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 2: &quot;Akopésáká&quot; (He will really give...)<br /> **Old output: &lt;code&gt;^Akopésáká/*Akopésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Akopésáká/kopésa&lt;v&gt;&lt;intns&gt;&lt;fut&gt;&lt;p3&gt;&lt;sg&gt;&lt;aa&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **Example sentence 3: &quot;Pésáká&quot; (Now do give...)<br /> **Old output: &lt;code&gt;^Pésáká/*Pésáká$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output &lt;code&gt;^Pésáká/kopésa&lt;v&gt;&lt;imp&gt;&lt;intns&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for more than 1 radical extension<br /> **Example sentence: &quot;Nabángisamákí na Pierre&quot; (I was scared by Pierre, or more literally, I was made, by Pierre, to fear something/someone)<br /> **Old output: &lt;code&gt;^Nabángisamákí/*Nabángisamákí$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^Nabángisamákí/kobánga&lt;v&gt;&lt;caus&gt;&lt;pass&gt;&lt;urp&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$ ^na/na&lt;pr&gt;/na&lt;cngcoo&gt;$ ^Pierre/Pierre&lt;n&gt;&lt;prop&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for numerals<br /> **Example sentence: &quot;zómi&quot; (ten)<br /> **Old output: &lt;code&gt;^zómi/*zómi$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^zómi/zómi&lt;num&gt;&lt;ord&gt;/zómi&lt;num&gt;&lt;card&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> *Support for quantifiers<br /> **Example sentence: &quot;mbelí nyónso&quot; (each knife)<br /> **Old output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/*nyónso$^./.&lt;sent&gt;$&lt;/code&gt;<br /> **New output: &lt;code&gt;^mbelí/mbelí&lt;n&gt;&lt;cl9&gt;&lt;sg&gt;/mbelí&lt;n&gt;&lt;cl9&gt;&lt;pl&gt;$ ^nyónso/nyónso&lt;det&gt;&lt;qnt&gt;$^./.&lt;sent&gt;$&lt;/code&gt;<br /> <br /> ====Additional Constraint Grammar Rules====<br /> *If I am an adjective and a quantifier but I do not agree in number with the preceding noun, I must be a quantifier<br /> **Example sentence: &quot;bilóko moké&quot; (A few things, &#039;&#039;not&#039;&#039; small things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> &quot;moké&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;moké&gt;&quot;<br /> &quot;moké&quot; det qnt<br /> ; &quot;moké&quot; adj sg REMOVE:21<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a singular adjective, I cannot be plural<br /> **Example sentence: &quot;nzelá molaí&quot; (a long road)<br /> **Old output (below) <br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> ; &quot;nzelá&quot; n cl9 pl REMOVE:25<br /> &quot;&lt;molaí&gt;&quot;<br /> &quot;molaí&quot; adj sg<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I am a noun that can be singular or plural and am followed by a plural adjective, I cannot be singular<br /> **Example sentence: &quot;nzelá milaí&quot; (long roads)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 sg<br /> &quot;nzelá&quot; n cl9 pl<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;nzelá&gt;&quot;<br /> &quot;nzelá&quot; n cl9 pl<br /> ; &quot;nzelá&quot; n cl9 sg REMOVE:28<br /> &quot;&lt;milaí&gt;&quot;<br /> &quot;molaí&quot; adj pl<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be an ordinal and a cardinal and am modifying a plural noun, I am not an ordinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;bilóko míbalé&quot; (two things, &#039;&#039;not&#039;&#039; the second thing)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> <br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;bilóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 pl<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num card<br /> ; &quot;míbalé&quot; num ord REMOVE:31<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be and ordinal and a cardinal and am modifying a a singular noun, I am not a cardinal. (One is never both an ordinal and cardinal, so this is not a problem.)<br /> **Example sentence: &quot;elóko mibalé&quot; (the second thing, &#039;&#039;not&#039;&#039; two things)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> &quot;míbalé&quot; num card<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;elóko&gt;&quot;<br /> &quot;elóko&quot; n cl7 sg<br /> &quot;&lt;míbalé&gt;&quot;<br /> &quot;míbalé&quot; num ord<br /> ; &quot;míbalé&quot; num card REMOVE:34<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> *If I can be a demonstrative nominal and a relative pronoun and am proceeded by a noun, I cannot be a demonstrative nominal.<br /> **Example sentence: &quot;Mobali óyo amonaki ngáí&quot; (The man who saw me...)<br /> **Old output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> &quot;óyo&quot; prn dem sg<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> :*New output (below)<br /> &lt;pre&gt;<br /> &quot;&lt;Mobali&gt;&quot;<br /> &quot;mobali&quot; n cl1 sg<br /> &quot;&lt;óyo&gt;&quot;<br /> &quot;óyo&quot; prn rel<br /> &quot;óyo&quot; det dem<br /> ; &quot;óyo&quot; prn dem sg REMOVE:37<br /> &quot;&lt;amonaki&gt;&quot;<br /> &quot;komóna&quot; v urp p3 sg aa<br /> &quot;&lt;ngáí&gt;&quot;<br /> &quot;ngáí&quot; prn pers p1 sg aa<br /> &quot;&lt;.&gt;&quot;<br /> &quot;.&quot; sent<br /> &lt;/pre&gt;<br /> <br /> ====Additional Transfer Rules====<br /> *Enhance existing basic verb transfer rule to work with newly implemented imperative<br /> **Example sentence: &quot;ina&quot; (sing)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;p1&gt;&lt;sg&gt;&lt;aa&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#koyemba&lt;/code&gt;<br /> **New output<br /> ***Tagger:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans:&lt;code&gt;^in&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;/koyemba&lt;v&gt;&lt;imp&gt;&lt;p1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk:&lt;code&gt;^verb&lt;verb&gt;{^koyemba&lt;v&gt;&lt;imp&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^koyemba&lt;v&gt;&lt;imp&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;yemba&lt;/code&gt;<br /> <br /> *Map Kikuyu associative to Lingala connective by removing extra agreement tags found in Kikuyu<br /> **Example sentence: &quot;wa&quot;<br /> **Old output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#ya&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^a&lt;assoc&gt;&lt;cl1&gt;&lt;sg&gt;/ya&lt;conn&gt;&lt;cl1&gt;&lt;sg&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^conn_chunk&lt;conn&gt;{^ya&lt;conn&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^ya&lt;conn&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;ya&lt;/code&gt;<br /> <br /> *Basic adjective transfer: keep number markings if a variable adjective in Lingala, discard otherwise<br /> **Example sentence 1: anene (big)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#monéne&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^nene&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;/monéne&lt;adj&gt;&lt;cl1&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^monéne&lt;adj&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^monéne&lt;adj&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;minéne&lt;/code&gt;<br /> **Example sentence 2: mata (sour)<br /> **Old output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^default&lt;default&gt;{^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;#bololo&lt;/code&gt;<br /> **New output<br /> ***Tagger: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Biltrans: &lt;code&gt;^mata&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;/bololo&lt;adj&gt;&lt;cl7&gt;&lt;pl&gt;$^.&lt;sent&gt;/.&lt;sent&gt;$&lt;/code&gt;<br /> ***Chunker: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Interchunk: &lt;code&gt;^adjective&lt;adj&gt;{^bololo&lt;adj&gt;$}$^sent&lt;SENT&gt;{^.&lt;sent&gt;$}$&lt;/code&gt;<br /> ***Postchunk: &lt;code&gt;^bololo&lt;adj&gt;$^.&lt;sent&gt;$&lt;/code&gt;<br /> ***Complete translation: &lt;code&gt;bololo&lt;/code&gt;<br /> <br /> ===lin Transducer Evaluation===<br /> *Precision (annotated corpus): 80.53%<br /> *Recall (annotated corpus): 55.64%<br /> *Coverage (large corpus): 62.10%<br /> *Number of words (large corpus): 275,546 (actual), 304,425 (tokenized)<br /> *Number of stems in transducer: 273<br /> <br /> ===kik Transducer Evaluation===<br /> *Precision (annotated corpus):<br /> *Recall (annotated corpus): <br /> *Coverage (large corpus):<br /> *Number of words (large corpus):<br /> *Number of stems in transducer:<br /> <br /> ===lin → kik MT Evaluation===<br /> *WER (longer corpus):<br /> *PER (longer corpus):<br /> *Percentage unknown words:<br /> *Percentage known words:<br /> *Trimmed coverage over longer corpus [lin-kik.automorf.bin applied to lin.longer.txt]: 89.29%<br /> *Trimmed coverage over large corpus [lin-kik.automorf.bin applied to lin.corpus.large.txt]: 48.90%<br /> *Number of stems in longer corpus: 117 (actual) 140 (tokenized, perhaps including punctuation)<br /> *Number of stems in large corpus: 275,546 (actual), 304,425 (tokenized)<br /> <br /> ===kik → lin MT Evaluation===<br /> *WER (longer corpus): 86.32%<br /> *PER (longer corpus): 75.21%<br /> *Percentage unknown words: 39.50% [Actually probably a bit higher since some words were marked with an &quot;@&quot; and not translated while the script only looks for words marked with a &quot;*&quot;]<br /> *Percentage known words: 60.60%<br /> *Trimmed coverage over longer corpus [kik-lin.automorf.bin applied to kik.corpus.larger.txt]:<br /> *Trimmed coverage over large corpus [kik-lin.automorf.bin applied to kik.longer.txt]:<br /> *Number of stems in longer corpus:<br /> *Number of stems in large corpus:<br /> <br /> <br /> [[Category:Sp17_TranslationPairs]]<br /> [[Category:Lingala]]<br /> [[Category:Kikuyu]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Lingala_and_Kikuyu/Lexical_Selection&diff=4081 2017-03-28T16:28:05Z

<p>Jmundo1: </p> <hr /> <div>This page details the lexical selection process for machine translation between Lingala and Kikuyu for one-to-many mappings.<br /> <br /> ==lin → kik==<br /> ===Solved Ambiguities===<br /> We have implemented rules to solve two ambiguous translations from Lingala to Kikuyu.<br /> ====Mabele====<br /> The Lingala noun mabele can be translated as follows:<br /> * (lin) mabele → (kik) thanga (sand); ria (milk)<br /> To (mostly) solve this, we made generalizations based on the meaning of the preceding rule. We chose a default translation of thanga since this is the only way mabele is used in our parallel corpus. We override the default translation when mabele is preceded by &quot;bololo&quot; (sour), &quot;kotoka&quot; (pour), or &quot;komela&quot; (drink), since these are the most likely contexts in which mabele would refer to milk. Obviously this misses a couple of cases, but it helps. The following sentences can be used to test both rules:<br /> * na mabele (&quot;from the earth/sand,&quot; should be translated thanga) (it&#039;s a bit odd to use mabele here, but this example is taken verbatim from our corpus)<br /> * nabuákí mabele (&quot;I throw sand,&quot; should be translated thanga)<br /> * bololo mabele (&quot;sour milk,&quot; should be translated ria)<br /> * atokí mabele (&quot;He pours milk,&quot; should be translated ria)<br /> * tomelí mabele (&quot;We drink milk,&quot; should be translated ria)<br /> <br /> ====Kobéta====<br /> The Lingala verb kobéta can be translated as follows: <br /> * (lin) kobéta → (kik) hũr (to beat (as in hit)); ur (to rain)<br /> <br /> ===Unsolved Ambiguities===<br /> ====Kokíma====<br /> The Lingala verb kokíma can be translated as follows:<br /> * (lin) kokíma → (kik) hanyũk (to run); ũr (to run away)<br /> This ambiguity is hard to solve because the difference between the two words is rather nuanced. One possible generalization is that running from something should be translated as running away (ũr), but this cannot be implemented because Lingala has a single preposition used for both from and to, among other meanings.<br /> ====Kofánda====<br /> The Lingala verb kofánda can be translated as follows:<br /> * (lin) kofánda → (kik) ikar (to sit); tũũr (to live (as in reside))<br /> This ambiguity is hard to solve because there is no obvious generalization that can be made. Both can be followed (and often are) by a preposition, so whether a preposition follows cannot be used in a rule. In general, the contexts are hard to distinguish. One might try to select the translation based on the object of the preposition following, if there is one, but this would end up being fairly complicated, and we had other examples of one-to-many mappings that could be more readily and thoroughly solved by lexical selection.<br /> ====kobánza====<br /> The Lingala verb kobánza can be translated as follows:<br /> * (lin) kobánza → (kik) ĩcir (to think); ririkan (to remember)<br /> This ambiguity is hard to solve mostly because we do not have a good understanding of how the syntax of kobánza works in Lingala, since we don&#039;t have many example sentences, and it is consequently hard to make any generalizations.<br /> ==kik → lin==<br /> ===Solved Ambiguities===<br /> We have implemented rules to solve one ambiguous translation from Kikuyu to Lingala.<br /> ====Ria====<br /> The Kikuyu noun ria can be translated as follows:<br /> * (kik) ria → (lin) mabele (milk); etimá (pond)<br /> To (mostly) solve this, we made generalizations based on the meaning of the preceding rule. We chose a default translation of etimá because it is easier to isolate situations where mabele is used than where etimá is used. We override the default translation when ria is followed by mata (thick/creamy), preceded or followed by it (pour/spill), preceded by (drink), or preceded by na (with)since these are likely contexts in which ria would refer to milk. As with the similar ambiguity in the other direction, this misses quite a few cases. The following phrases can be used to test the rules<br /> * iria imata (&quot;thick milk&quot;, should be translated mabele)<br /> * na iria (&quot;with milk&quot;, should be translated mabele)<br /> * nĩnyuire iria (&quot;I drank milk&quot;, should be translated mabele)<br /> * nĩnjitire iria (&quot;I poured the milk&quot;, should be translated mabele)<br /> * iria nĩrĩitĩkire (&quot;The milk spilled&quot;, should be translated mabele)<br /> * iriainĩ (sometimes spelled iria-inĩ) (&quot;in the lake&quot;, should be translated etímá)<br /> * iria inene (&quot;the big lake&quot;, should be translated etímá)<br /> ===Unsolved Ambiguities===<br /> ====Igu====<br /> The Kikuyu verb igu can be translated as follows:<br /> * (kik) igu → (lin) kotosa (to obey); koyoka (to hear)<br /> This ambiguity is hard to solve because many of the contexts in which the word for hear might appear are also contexts in which obey might appear and vice versa. For example, although one might obey a command, one might also hear a command.<br /> ====Rut====<br /> The Kikukyu verb rut can be translated as follows:<br /> * (kik) rut → (lin) koyekola (to learn); kokukula (to remove)<br /> This ambiguity is hard to solve because, although the two meanings are far removed from each other, the contextual information that lets a hearer select the correct meaning is much more sophisticated than a small set of lemmas or tags. It is therefore hard to make a generalization that could be implemented as a rule.<br /> <br /> <br /> [[Category:Sp17_LexicalSelection]]<br /> [[Category:Lingala]]<br /> [[Category:Kikuyu]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Lingala_and_Kikuyu/Lexical_Selection&diff=4073 2017-03-28T16:18:00Z

<p>Jmundo1: </p> <hr /> <div>This page details the lexical selection process for machine translation between Lingala and Kikuyu for one-to-many mappings.<br /> <br /> ==lin → kik==<br /> ===Solved Ambiguities===<br /> We have implemented rules to solve two ambiguous translations from Lingala to Kikuyu.<br /> ====Mabele====<br /> The Lingala noun mabele can be translated as follows:<br /> * (lin) mabele → (kik) thanga (sand); ria (milk)<br /> To (mostly) solve this, we made generalizations based on the meaning of the preceding rule. We chose a default translation of thanga since this is the only way mabele is used in our parallel corpus. We override the default translation when mabele is preceded by &quot;bololo&quot; (sour), &quot;kotoka&quot; (pour), or &quot;komela&quot; (drink), since these are the most likely contexts in which mabele would refer to milk. Obviously this misses a couple of cases, but it helps. The following sentences can be used to test both rules:<br /> * na mabele (&quot;from the earth/sand,&quot; should be translated thanga) (it&#039;s a bit odd to use mabele here, but this example is taken verbatim from our corpus)<br /> * nabuákí mabele (&quot;I throw sand,&quot; should be translated thanga)<br /> * bololo mabele (&quot;sour milk,&quot; should be translated ria)<br /> * atokí mabele (&quot;He pours milk,&quot; should be translated ria)<br /> * tomelí mabele (&quot;We drink milk,&quot; should be translated ria)<br /> <br /> ====Kobéta====<br /> The Lingala verb kobéta can be translated as follows: <br /> * (lin) kobéta → (kik) hũr (to beat (as in hit)); ur (to rain)<br /> <br /> ===Unsolved Ambiguities===<br /> ====Kokíma====<br /> The Lingala verb kokíma can be translated as follows:<br /> * (lin) kokíma → (kik) hanyũk (to run); ũr (to run away)<br /> <br /> ====Kofánda====<br /> The Lingala verb kofánda can be translated as follows:<br /> * (lin) kofánda → (kik) ikar (to sit); tũũr (to live (as in reside))<br /> <br /> ====kobánza====<br /> The Lingala verb kobánza can be translated as follows:<br /> * (lin) kobánza → (kik) ĩcir (to think); ririkan (to remember)<br /> <br /> <br /> ==kik → lin==<br /> ===Solved Ambiguities===<br /> We have implemented rules to solve one ambiguous translation from Kikuyu to Lingala.<br /> ====Ria====<br /> The Kikuyu noun ria can be translated as follows:<br /> * (kik) ria → (lin) mabele (milk); etimá (pond)<br /> To (mostly) solve this, we made generalizations based on the meaning of the preceding rule. We chose a default translation of etimá because it is easier to isolate situations where mabele is used than where etimá is used. We override the default translation when ria is followed by mata (thick/creamy), preceded or followed by it (pour/spill), preceded by (drink), or preceded by na (with)since these are likely contexts in which ria would refer to milk. As with the similar ambiguity in the other direction, this misses quite a few cases. The following phrases can be used to test the rules<br /> * iria imata (&quot;thick milk&quot;, should be translated mabele)<br /> * na iria (&quot;with milk&quot;, should be translated mabele)<br /> * nĩnyuire iria (&quot;I drank milk&quot;, should be translated mabele)<br /> * nĩnjitire iria (&quot;I poured the milk&quot;, should be translated mabele)<br /> * iria nĩrĩitĩkire (&quot;The milk spilled&quot;, should be translated mabele)<br /> * iriainĩ (sometimes spelled iria-inĩ) (&quot;in the lake&quot;, should be translated etímá)<br /> * iria inene (&quot;the big lake&quot;, should be translated etímá)<br /> ===Unsolved Ambiguities===<br /> ====Igu====<br /> The Kikuyu verb igu can be translated as follows:<br /> * (kik) igu → (lin) kotosa (to obey); koyoka (to hear)<br /> ====Rut====<br /> The Kikukyu verb rut can be translated as follows:<br /> * (kik) rut → (lin) koyekola (to learn); kokukula (to remove)<br /> <br /> <br /> <br /> [[Category:Sp17_LexicalSelection]]<br /> [[Category:Lingala]]<br /> [[Category:Kikuyu]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Lingala_and_Kikuyu/Lexical_Selection&diff=4069 2017-03-28T16:12:36Z

<p>Jmundo1: </p> <hr /> <div>This page details the lexical selection process for machine translation between Lingala and Kikuyu for one-to-many mappings.<br /> <br /> ==lin → kik==<br /> ===Solved Ambiguities===<br /> We have implemented rules to solve two ambiguous translations from Lingala to Kikuyu.<br /> ====Mabele====<br /> The Lingala noun mabele can be translated as follows:<br /> * (lin) mabele → (kik) thanga (sand); ria (milk)<br /> To (mostly) solve this, we made generalizations based on the meaning of the preceding rule. We chose a default translation of thanga since this is the only way mabele is used in our parallel corpus. We override the default translation when mabele is preceded by &quot;bololo&quot; (sour), &quot;kotoka&quot; (pour), or &quot;komela&quot; (drink), since these are the most likely contexts in which mabele would refer to milk. Obviously this misses a couple of cases, but it helps. The following sentences can be used to test both rules:<br /> * na mabele (&quot;from the earth/sand,&quot; should be translated thanga) (it&#039;s a bit odd to use mabele here, but this example is taken verbatim from our corpus)<br /> * nabuákí mabele (&quot;I throw sand,&quot; should be translated thanga)<br /> * bololo mabele (&quot;sour milk,&quot; should be translated ria)<br /> * atokí mabele (&quot;He pours milk,&quot; should be translated ria)<br /> * tomelí mabele (&quot;We drink milk,&quot; should be translated ria)<br /> <br /> ====Kobéta====<br /> The Lingala verb kobéta can be translated as follows: <br /> * (lin) kobéta → (kik) hũr (to beat (as in hit)); ur (to rain)<br /> <br /> ===Unsolved Ambiguities===<br /> ====Kokíma====<br /> The Lingala verb kokíma can be translated as follows:<br /> * (lin) kokíma → (kik) hanyũk (to run); ũr (to run away)<br /> <br /> ====Kofánda====<br /> The Lingala verb kofánda can be translated as follows:<br /> * (lin) kofánda → (kik) ikar (to sit); tũũr (to live (as in reside))<br /> <br /> ====kobánza====<br /> The Lingala verb kobánza can be translated as follows:<br /> * (lin) kobánza → (kik) ĩcir (to think); ririkan (to remember)<br /> <br /> <br /> ==kik → lin==<br /> ===Solved Ambiguities===<br /> We have implemented rules to solve one ambiguous translation from Kikuyu to Lingala.<br /> ====Ria====<br /> The Kikuyu noun ria can be translated as follows:<br /> * (kik) ria → (lin) mabele (milk); etimá (pond)<br /> To (mostly) solve this, we made generalizations based on the meaning of the preceding rule. We chose a default translation of etimá because it is easier to isolate situations where mabele is used than where etimá is used. We override the default translation when ria is followed by mata (thick/creamy), preceded or followed by it (pour/spill), preceded by (drink), or preceded by na (with)since these are likely contexts in which ria would refer to milk. As with the similar ambiguity in the other direction, this misses quite a few cases. The following phrases can be used to test the rules<br /> * iria imata (&quot;thick milk&quot;, should be translated mabele)<br /> * na iria (&quot;with milk&quot;, should be translated mabele)<br /> * nĩnyuire iria (&quot;I drank milk&quot;, should be translated mabele)<br /> * nĩnjitire iria (&quot;I poured the milk&quot;, should be translated mabele)<br /> * iria nĩrĩitĩkire (&quot;The milk spilled&quot;, should be translated mabele)<br /> * iriainĩ (sometimes spelled iria-inĩ) (&quot;in the lake&quot;, should be translated etimá)<br /> ===Unsolved Ambiguities===<br /> ====Igu====<br /> The Kikuyu verb igu can be translated as follows:<br /> * (kik) igu → (lin) kotosa (to obey); koyoka (to hear)<br /> ====Rut====<br /> The Kikukyu verb rut can be translated as follows:<br /> * (kik) rut → (lin) koyekola (to learn); kokukula (to remove)<br /> <br /> <br /> <br /> [[Category:Sp17_LexicalSelection]]<br /> [[Category:Lingala]]<br /> [[Category:Kikuyu]]</div>

Jmundo1 http://wikis.swarthmore.edu/ling073/index.php?title=Lingala_and_Kikuyu/Lexical_Selection&diff=4066 2017-03-28T16:10:22Z

<p>Jmundo1: </p> <hr /> <div>This page details the lexical selection process for machine translation between Lingala and Kikuyu for one-to-many mappings.<br /> <br /> ==lin → kik==<br /> ===Solved Ambiguities===<br /> We have implemented rules to solve two ambiguous translations from Lingala to Kikuyu.<br /> ====Mabele====<br /> The Lingala noun mabele can be translated as follows:<br /> * (lin) mabele → (kik) thanga (sand); ria (milk)<br /> To (mostly) solve this, we made generalizations based on the meaning of the preceding rule. We chose a default translation of thanga since this is the only way mabele is used in our parallel corpus. We override the default translation when mabele is preceded by &quot;bololo&quot; (sour), &quot;kotoka&quot; (pour), or &quot;komela&quot; (drink), since these are the most likely contexts in which mabele would refer to milk. Obviously this misses a couple of cases, but it helps. The following sentences can be used to test both rules:<br /> * na mabele (&quot;from the earth/sand,&quot; should be translated thanga) (it&#039;s a bit odd to use mabele here, but this example is taken verbatim from our corpus)<br /> * nabuákí mabele (&quot;I throw sand,&quot; should be translated thanga)<br /> * bololo mabele (&quot;sour milk,&quot; should be translated ria)<br /> * atokí mabele (&quot;He pours milk,&quot; should be translated ria)<br /> * tomelí mabele (&quot;We drink milk,&quot; should be translated ria)<br /> <br /> ====Kobéta====<br /> The Lingala verb kobéta can be translated as follows: <br /> * (lin) kobéta → (kik) hũr (to beat (as in hit)); ur (to rain)<br /> <br /> ===Unsolved Ambiguities===<br /> ====Kokíma====<br /> The Lingala verb kokíma can be translated as follows:<br /> * (lin) kokíma → (kik) hanyũk (to run); ũr (to run away)<br /> <br /> ====Kofánda====<br /> The Lingala verb kofánda can be translated as follows:<br /> * (lin) kofánda → (kik) ikar (to sit); tũũr (to live (as in reside))<br /> <br /> ====kobánza====<br /> The Lingala verb kobánza can be translated as follows:<br /> * (lin) kobánza → (kik) ĩcir (to think); ririkan (to remember)<br /> <br /> <br /> ==kik → lin==<br /> ===Solved Ambiguities===<br /> We have implemented rules to solve one ambiguous translation from Kikuyu to Lingala.<br /> ====Ria====<br /> The Kikuyu noun ria can be translated as follows:<br /> * (kik) ria → (lin) mabele (milk); etimá (pond)<br /> To (mostly) solve this, we made generalizations based on the meaning of the preceding rule. We chose a default translation of etimá because it is easier to isolate situations where mabele is used than where etimá is used. We override the default translation when ria is followed by mata (thick/creamy), preceded or followed by it (pour/spill), preceded by (drink), or preceded by na (with)since these are likely contexts in which ria would refer to milk. As with the similar ambiguity in the other direction, this misses quite a few cases. The following phrases can be used to test the rules<br /> * iria imata (&quot;thick milk&quot;, should be translated mabele)<br /> * na iria (&quot;with milk&quot;, should be translated mabele)<br /> * nĩnyuire iria (&quot;I drank milk&quot;, should be translated mabele)<br /> * nĩnjitire iria (&quot;I poured the milk&quot;, should be translated mabele)<br /> * iria nĩrĩitĩkire (&quot;The milk spilled&quot;, should be translated mabele -- note this translation will not actually work yet because the Kikuyu transducer is not yet equipped to analyse the middle voice)<br /> * iriainĩ (sometimes spelled iria-inĩ) (&quot;in the lake&quot;, should be translated etimá)<br /> ===Unsolved Ambiguities===<br /> ====Igu====<br /> The Kikuyu verb igu can be translated as follows:<br /> * (kik) igu → (lin) kotosa (to obey); koyoka (to hear)<br /> ====Rut====<br /> The Kikukyu verb rut can be translated as follows:<br /> * (kik) rut → (lin) koyekola (to learn); kokukula (to remove)<br /> <br /> <br /> <br /> [[Category:Sp17_LexicalSelection]]<br /> [[Category:Lingala]]<br /> [[Category:Kikuyu]]</div>

Jmundo1