Difference between revisions of "Judeo-Tat/Disambiguation"
(→Initial evaluation of ambiguity) |
(→Final evaluation of ambiguity) |
||
| (8 intermediate revisions by the same user not shown) | |||
| Line 1: | Line 1: | ||
| + | ==Final evaluation of ambiguity== | ||
| + | After implementing rules and fixing pronouns, there is an ambiguity of ~1.167. | ||
| + | |||
| + | 99 out of 102 tests are now passing in <code>jdt.yaml</code> | ||
| + | |||
| + | The word I focused on is "е", which can be either an adverb or an article. | ||
| + | |||
| + | After adding this word and several other words needed to full define the sentences I was using for disambiguation, I have a corpus coverage of 37.07%. | ||
| + | |||
| + | I made my rules based on whether or not the word after "е" was a classifier or not. If it was, then "е" was an article, and if it wasn't, it was an adverb. | ||
| + | |||
| + | After implementing the rules, this is what I had for my two example sentences. | ||
| + | |||
| + | The first should have "е" as an article. | ||
| + | |||
| + | <code>echo "ме е доне гендуьме егужевоз фуберденуьм" | apertium -d . jdt-disam</code> | ||
| + | |||
| + | "I had trouble swallowing a single grain of wheat" | ||
| + | <pre> | ||
| + | "<ме>" | ||
| + | "ме" prn pers p1 sg | ||
| + | "<е>" | ||
| + | "е" art p1 | ||
| + | ; "е" adv REMOVE:9 | ||
| + | "<доне>" | ||
| + | "доне" class | ||
| + | "<гендуьме>" | ||
| + | "гендуьм" n dat | ||
| + | "<егужевоз>" | ||
| + | "гуж" n loc ins | ||
| + | "<фуберденуьм>" | ||
| + | "фубер" n prs p1 | ||
| + | "<.>" | ||
| + | "." sent | ||
| + | </pre> | ||
| + | And the second should have "е" as an adverb. | ||
| + | |||
| + | <code>echo "ез егеми вебируьтгьо одомигьо е у согъ мунд" | apertium -d . jdt-disam | ||
| + | </code> | ||
| + | |||
| + | "Of those who were on the ship, he alone remained alive" | ||
| + | |||
| + | <pre> | ||
| + | "<ез>" | ||
| + | "ез" abl | ||
| + | "<егеми>" | ||
| + | "геми" n loc | ||
| + | "<вебируьтгьо>" | ||
| + | "вебируьт" vblex past p3 pl | ||
| + | "<одомигьо>" | ||
| + | "одоми" n pl | ||
| + | "<е>" | ||
| + | "е" adv | ||
| + | ; "е" art p1 REMOVE:12 | ||
| + | "<у>" | ||
| + | "у" prn dem sg nom | ||
| + | "<согъ>" | ||
| + | "согъ" adj | ||
| + | "<мунд>" | ||
| + | "мун" vblex aor sg p3 | ||
| + | "<.>" | ||
| + | "." sent | ||
| + | |||
| + | </pre> | ||
| + | |||
==Initial evaluation of ambiguity== | ==Initial evaluation of ambiguity== | ||
| − | There is initially an ambiguity of ~1.308. All of my pronouns were ambiguous. | + | There is initially an ambiguity of ~1.308. |
| + | |||
| + | All of my pronouns were ambiguous, which I fixed. | ||
* {{morphTest|ме{{tag|prn}}{{tag|pers}}{{tag|p1}}{{tag|sg}}|ме}} | * {{morphTest|ме{{tag|prn}}{{tag|pers}}{{tag|p1}}{{tag|sg}}|ме}} | ||
* {{morphTest|туь{{tag|prn}}{{tag|pers}}{{tag|p2}}{{tag|sg}}|туь}} | * {{morphTest|туь{{tag|prn}}{{tag|pers}}{{tag|p2}}{{tag|sg}}|туь}} | ||
| Line 11: | Line 78: | ||
* {{morphTest|и{{tag|prn}}{{tag|dem}}{{tag|pl}}|игьо}} | * {{morphTest|и{{tag|prn}}{{tag|dem}}{{tag|pl}}|игьо}} | ||
* {{morphTest|у{{tag|prn}}{{tag|dem}}{{tag|pl}}|угьо}} | * {{morphTest|у{{tag|prn}}{{tag|dem}}{{tag|pl}}|угьо}} | ||
| − | |||
| − | |||
[[Category:Judeo-Tat]][[Category:Sp17_Disambiguation]] | [[Category:Judeo-Tat]][[Category:Sp17_Disambiguation]] | ||
Latest revision as of 17:31, 28 February 2017
Final evaluation of ambiguity
After implementing rules and fixing pronouns, there is an ambiguity of ~1.167.
99 out of 102 tests are now passing in jdt.yaml
The word I focused on is "е", which can be either an adverb or an article.
After adding this word and several other words needed to full define the sentences I was using for disambiguation, I have a corpus coverage of 37.07%.
I made my rules based on whether or not the word after "е" was a classifier or not. If it was, then "е" was an article, and if it wasn't, it was an adverb.
After implementing the rules, this is what I had for my two example sentences.
The first should have "е" as an article.
echo "ме е доне гендуьме егужевоз фуберденуьм" | apertium -d . jdt-disam
"I had trouble swallowing a single grain of wheat"
"<ме>" "ме" prn pers p1 sg "<е>" "е" art p1 ; "е" adv REMOVE:9 "<доне>" "доне" class "<гендуьме>" "гендуьм" n dat "<егужевоз>" "гуж" n loc ins "<фуберденуьм>" "фубер" n prs p1 "<.>" "." sent
And the second should have "е" as an adverb.
echo "ез егеми вебируьтгьо одомигьо е у согъ мунд" | apertium -d . jdt-disam
"Of those who were on the ship, he alone remained alive"
"<ез>" "ез" abl "<егеми>" "геми" n loc "<вебируьтгьо>" "вебируьт" vblex past p3 pl "<одомигьо>" "одоми" n pl "<е>" "е" adv ; "е" art p1 REMOVE:12 "<у>" "у" prn dem sg nom "<согъ>" "согъ" adj "<мунд>" "мун" vblex aor sg p3 "<.>" "." sent
Initial evaluation of ambiguity
There is initially an ambiguity of ~1.308.
All of my pronouns were ambiguous, which I fixed.
- ме<prn><pers><p1><sg> ↔ ме
- туь<prn><pers><p2><sg> ↔ туь
- иму<prn><pers><p1><pl> ↔ иму
- ишму<prn><pers><p2><pl> ↔ ишму
- и<prn><dem><sg><nom> ↔ и
- у<prn><dem><sg><nom> ↔ у
- и<prn><dem><sg> ↔ ире
- у<prn><dem><sg> ↔ уре
- и<prn><dem><pl> ↔ игьо
- у<prn><dem><pl> ↔ угьо
