APPLICATION-ORIENTED COMPARISON AND EVALUATION OF SIX SEMANTIC SIMILARITY MEASURES BASED ON WORDNET

摘要

In the task of auto-building a Chinese-English semantic lexicon for translation selection, this research presents a method, which introduces WordNet similarity measures to wash out misaligned Chinese-English word pairs. Six different proposed measures of similarity based on WordNet were experimentally compared and evaluated by using WordNet and the software package WordNet::Similarity1. It was found that the leader measure is res and lch, then wup, lin takes the fourth place, thenjcn, and random is the tailender.