Difference between revisions of "Guarani and Warlpiri"

From LING073
Jump to: navigation, search
(→‎Disambiguation)
(→‎Morphology additions)
Line 120: Line 120:
  
 
===Morphology additions===
 
===Morphology additions===
 +
*Improved pronomial analysis:
 +
*Before:
 +
**"<ngaju>"
 +
***"ngaju" prn pers p3 sg
 +
***"ngaju" prn pers p3 du
 +
***"ngaju" prn pers p2 sg
 +
***"ngaju" prn pers p2 du
 +
***"ngaju" prn pers p2 pl
 +
***"ngaju" prn pers p1 sg
 +
***"ngaju" prn pers p1 du excl
 +
***"ngaju" prn pers p1 du incl
 +
***"ngaju" prn pers p1 pl excl
 +
***"ngaju" prn pers p1 pl incl
 +
*Before:
 +
**"<ngaju>"
  
 +
 +
-----------------------------------
 
*Clitics
 
*Clitics
 
** (wbp) -yijala  cl. also, too, likewise.
 
** (wbp) -yijala  cl. also, too, likewise.

Revision as of 00:30, 18 April 2017

Resources for machine translation between Guarani and Warlpiri

Corpora

Developed Resources

Lexical Selection

  • (gug) ahy'o → (wbp) ngukarna "throat"; lirra "voice"
  • (gug) ha'anga → (wbp) japirdimi "threaten"; walaparrirni "imitate"
  • (gug) ajúra → (wbp) japirdimi "collar"; waninja "neck"
  • (wbp) rdaka → (gug) avatisoka "hand"; po "five"
  • (wbp) kirdirrpa → (gug) itakua "cave"; ka'irãi "jail"

Initial Evaluation

wbp → gug evaluation

Results when running on tests

  • Total number of words: 14
  • Total number of unknown words: 4
  • Percentage of unknown words: 28.57%
  • Percentage of stems translated correctly: 71.43%

When removing unknown words:

  • WER = 78.57%
  • PER = 78.57%

When unknown words are not removed:

  • WER = 100%
  • PER = 100%

Results when running on sentences (corpus)

  • Total number of words: 58
  • Total number of unknown words: 68
  • Percentage of unknown words: 84.45%
  • Percentage of stems translate correctly: 15.52%

When removing unknown words:

  • WER = 22.41%
  • PER = 22.41%

When unknown words are not removed:

  • WER = 100%
  • PER = 100%


gug → wbp evaluation

Results when running on tests

  • Total number of words: 13
  • Total number of unknown words: 0
  • Percentage of unknown words: 0%
  • Percentage of stems translated correctly: 100%

When removing unknown words:

  • WER = 92.86%
  • PER = 92.86%

When unknown words are not removed:

  • WER = 92.86%
  • PER = 92.86%

Results when running on sentences (corpus)

  • Total number of words: 66
  • Total number of unknown words: 60
  • Percentage of unknown words: 90.9%
  • Percentage of stems translated correctly: 9.09%

When removing unknown words:

  • WER = 0.0%
  • PER = 0.0%

When unknown words are not removed:

  • WER = 100%
  • PER = 100%

Final Evaluation

100 more stems

Added 100 more stems (mostly nouns, verbs, and adjectives) to the bilingual and monolingual dictionaries.

3 new lexical selection rules

gug → wbp

  • (gug) ahy'o → (wbp) ngukarna "throat"; lirra "voice"
    • Rule: Select "lirra" (Eng: voice) as the translation of "ahy'o" when it is followed by "tyapu" (Eng: loud) or "kirirĩ" (Eng: quiet). Select "ngukarna" (Eng: throat) as the translation of "ahy'o" when it is followed by "rasy" (Eng: sore). Otherwise, select "lirra" as the default.
    • Ha'e ahy'o tyapu ("His/Her loud voice")
    • Ha'e oguereko ahy'o rasy ("He/she/they have a sore throat")
  • (gug) ajúra → (wbp) japirdimi "collar"; waninja "neck"
    • Rule: Select "japirdimi" (Eng: collar) as the translation of "ajúra" when it is followed by color words such as "pytã" (Eng: red), "chovy" (Eng: blue) or "aky" (Eng: green). Select "waninja" (Eng: neck) as the translation of "ajúra" when it is followed by "puku" (Eng: long) or "koli" (Eng: short). Otherwise, select "waninja" as the default.
    • Ajúra puku ("long neck")
    • Ajúra pytã ("red collar")
  • (gug) akã → (wbp) jurru "head"; payarrpa "hill"
  • (gug) asu → (wbp) jampu "left"; jampu-karra "left-handed"

wbp → gug

  • (wbp) muru-pinyi → (gug) mboy'u "swallow"; kuapyti "insert"
    • Rule: select "mboy'u" if "muru-pinyi" is followed by food or drink words: "ngapa" (water/grog), "pama" (liquor), "jirnarri" (food), and "mangarri" (vegetable). In other cases, select "kuapyti".
    • Kurdu-ngku muru-pinyi ngapa. (Eng: The child is swallowing water/grog.)
    • muru-pinyi walya (Eng: insert in the ground)
  • (wbp) kirdirrpa → (gug) itakua "cave"; ka'irãi "jail"
    • Rule: We couldn't come up with a good way to solve this problem. We decided to select "itakua" as the default, because it appears to be more common.
    • Kurdu-ngku nyanyi kirdirrpa. (Eng: The child is looking for the cave.)
    • Wati-ngku jarda-jarrimi kirdirrpa-rla (Eng: The man sleeps in the jail.)


  • (wbp) juka → (gug) asuka "sugar"; aichémemby "cousin"
  • (wbp) warlu → (gug) tini "hot"; aku "angry"
  • (wbp) yajarni → (gug) ? "fetch/invite (someone)"; aku "grow (a plant)"

Morphology additions

  • Improved pronomial analysis:
  • Before:
    • "<ngaju>"
      • "ngaju" prn pers p3 sg
      • "ngaju" prn pers p3 du
      • "ngaju" prn pers p2 sg
      • "ngaju" prn pers p2 du
      • "ngaju" prn pers p2 pl
      • "ngaju" prn pers p1 sg
      • "ngaju" prn pers p1 du excl
      • "ngaju" prn pers p1 du incl
      • "ngaju" prn pers p1 pl excl
      • "ngaju" prn pers p1 pl incl
  • Before:
    • "<ngaju>"


  • Clitics
    • (wbp) -yijala cl. also, too, likewise.
    • (wbp) -nyayirni Variant: -nyarrirni. cl. very, really -> (gug) adverb: asy
  • Conjunction??
    • ngula cj. 1. that, which, who. Syn: kuja. 2.he, she, it, they, them. Referring back to something in previous context.
  • Cs?
    • -jangka cs. 1. from. 2. because. Syn: -ngurlu~-ngirli, -warnu.
  • Reflexive?
    • nyanungu pn.
  • ngana - who
  • yangka nm. aforementioned one, same one mentioned before.
  • Numbers
  • -juku cl. still, yet. Syn: -kirli.
  • -pinki ns. such-like, things pertaining to.
  • -kari ns. another, other.
  • nyurru av. already, short time ago.
  • -wiyi (first)
  • -wana ns. along, alongside, beside.

Disambiguation

  • Number vs Noun (e.g. rdaka = hand or five), has the number marker -pala
  • Manu = cnjcoo vs v tv past?
Retrieved from "http://wikis.swarthmore.edu/ling073/index.php?title=Guarani_and_Warlpiri&oldid=4583"