Two statistics methods of Chinese word sense disambiguation

Word Sense Disambiguation (WSD) has always been a difficult and hot points in natural language processing. At present, only some ambiguous words are selected as research objects in most WSD research, which has large gap with the real application. In this paper, large scale real texts are applied in WSD based on two classical statistics model. The supervised WSD method based on Hidden Markov Model (HMM) got a lower precision, only about 85% in open test. The precision of the method based on Naive Bayes Model (NBM) is 92%, it's a higher precision. And the unsupervised WSD based on NBM got a little lower precision in comparison to the supervised, but it is worthy further researching since it has a well extension performance.