An Approach Based on Tongyici Cilin and Word Similarity for Chinese Word Sense Induction

摘要

This paper presents a new approach of automatic unsupervised Word Sense Induction. This approach is based on Tongyici Cilin (a Chinese synonym dictionary). First we extract the neighbor words with consideration of the POS tags. Then, we calculate the word similarity according to the semantic code mapped in Tongyici Cilin (Extended). Finally, we design a greedy algorithm to accomplish clustering. The experimental results indicate that our approach is very potentially promising based on the benchmark data set provided by CIPS and SIGHAN. The work has an important significance for the usage of this thesaurus in Word Similarity Calculation and Word Sense Disambiguation.