Augmenting a Bilingual Lexicon with Information for Word Translation Disambiguation

Takashi Tsunakawa, Hiroyuki Kaji · 2010

We describe a method for augmenting a bilingual lexicon with additional information for selecting an appropriate translation word. For each word in the source language, we calculate a correlation matrix of its association words versus its translation candidates. We estimate the degree of correlation by using comparable corpora based on these assumptions: “parallel word associations” and “one sense per word association. ” In our word translation disambiguation experiment, the results show that our method achieved 42% recall and 49 % precision for Japanese-English newspaper texts, and 45% recall and 76 % precision for Chinese-Japanese technical documents. 1

Read the paper · More papers on PaperTik