Difference between revisions of "Walloon and English/AmbiguityResolution"

Revision as of 17:22, 27 April 2025

Initial Evaluation of Ambiguity

• Parallel sentences (wln.sentences.txt)
• Tokens – 86
• Analyses – 91
• Ambiguity – 1.06

• Largest monolingual corpus (wln.corpus.basic.txt)
• Tokens – 1,356
• Analyses – 1,381
• Ambiguity – 1.018

Identified Ambiguity

Morphological ambiguity

• wln: Dji houke mes amîs.
• analysis: ^Dji/dji<prn><p1><sg>$ ^houke/houker<v><pres><p1><sg>/houker<v><pres><p3><sg>$
• eng: I call my friends.

• wln: Djezus houke sès prumîs dissipes.
• analysis: ^Djezus/Djezus<np><ant><m><sg>$ ^houke/houker<v><pres><p1><sg>/houker<v><pres><p3><sg>$
• eng: Jesus calls his first disciples.

• Rule: when a 1 sg pronoun (dji) precedes the verb, select the p1 reading, when a 3 sg pronoun (i / l') precedes, or no cue, select p3.

• Extra picks between s' <det><poss> and <prn> by looking at the following chunk (noun -> <det><poss>; verb -> <prn>).

Lexical ambiguity

• wln: Il vindrè di Liège.
• analysis: ^Il/il<prn><p3><sg>$ ^vindrè/vindr{ER}<v><fut><p3><sg>$ ^di/di<pr>$
• eng: He will come from Liège.

• wln: On djoû di djuillet.
• analysis: ^On/on<det><ind><m><sg>$ ^djoû/djoû<n><m><sg>$ ^di/di<pr>$
• eng: A day of July.

• Rule: if verb is immediately before di, translate as from; otherwise default to of.

Final Evaluation of Ambiguity

• Parallel sentences - analyses: 86 / tokens: 86, 1.00
• Largest monolingual corpus - analyses: 1,357 / tokens: 1,356, 1.001