Difference between revisions of "Fijian/Disambiguation"

Revision as of 16:12, 20 March 2018

Initial Evaluation of Ambiguity

• Current coverage after adding some more common words: 48.41%.
• Initial ambiguity: 1558/1408=1.106

Ambiguous Analyses

• Sentence 1: era sa la’o o ira. (not in the current corpus)

-Analysis: ^era/ra<prn><pers><pl><p3><subj1>$ ^sa/sa<asp>$ ^la’o/la’o<va><iv>$ ^o/o<art>/o<prn><pers><sg><p2><subj>$ ^ira/ra<prn><pers><pl><p3><obj>/ra<prn><pers><pl><p3><card1>$

• Sentence 2: ’Eimami aa la’o mai Suva.

-Analysis: ^’Eimami/imami<prn><pers><pl><p1><exc><subj>/imami<prn><pers><pl><p1><exc><cl2_pos>/imami<prn><pers><pl><p1><exc><obj>/imami<prn><pers><pl><p1><exc><card>$ ^aa/aa<asp>$ ^la’o/la’o<va><iv>$ ^mai/mai<pr>$ ^Suva/Suva<top>$

Constraints

• For ambiguity 1: If a pronoun is preceded by an article, then it is in the cardinal form.
• For ambiguity 2: If a pronoun is followed by verbs, then it is in the subject form.

After Disambiguation

• Current coverage: 51.09%
• Sentence 1: The constraint depends on the environment preceding the pronoun. How to implement this in the rlx file?
• Sentence 2:

$ echo "’Eimami aa la’o mai Suva" | apertium -d . fij-tagger

^imami<prn><pers><pl><p1><exc><subj>$ ^aa<asp>$ ^la’o<va><iv>$ ^mai<pr>$ ^Suva<top>$^.<sent>$

• Remaining problems:

-The $ echo "’Eimami aa la’o mai Suva" | apertium -d . fij-disam command does not show any disambiguation.

-Sentence 1.