A hybrid approach for Chinese word similarity computing based on HowNet

Word similarity has a wide range of applications, such as machine translation, word sense disambiguation ^{[1, 5]}, text classification, information extraction and so on. HowNet is an online common-sense knowledge system ^[12]. It´s described by a language called KDML (Knowledge Database Markup Language). In this paper, we use the resources provided by HowNet 2002 ^[11] to compute the semantic similarity of words, and propose a hybrid approach for calculating word similarity using tree-similarity and Markov model.