Chinese word sense induction based on topic model

Abstract:

Sense Induction is the process of identifying the word sense given its context, often treated as a clustering task. In this paper, we present a approach to Chinese Word Sense Induction which is based on topic modeling. Key to our methodology is the use of probabilistic assignment of topics distributions to documents to estimate sense distributions. Experimental results show that our method could achieve 77.58% scores of F-score on the development data set.