An Unsupervised Approach To Word Sense Disambiguation Based On Second-order Context

Word sense disambiguation (WSD) is always a key problem in natural language processing because the result of WSD affects seriously many problems in nature language processing and information retrieval. WSD is essential for many language applications such as machine translation, information retrieval, natural language semantic analysis, grammar analysis, speech identification, and conversion from text to speech and so on, What is more, it has great theory and reality significance on realizing and grasping the actuality and developing trends.Now, some foreign researchers and domain researchers have done lots of study on word sense disambiguation based on supervised and unsupervised approaches. They do experiments by means of all kinds of dictionaries and get satisfy results. Because supervised WSD need to label the trained corpus, which cost lots of time and labor and there is serious sparseness of data in the statistical result, so many researchers dedicate in unsupervised knowledge learning.This paper gives details to the research on unsupervised approach. The main aspects of the paper are as follows:(1) We proposed an unsupervised approach based on digital dictionary. We can do word sense disambiguation for many contexts of one Polysemous word at the same time;(2) We classify the contexts of the Polysemous word by clustering of k-means and the contexts in each category correspond to the same word sense;(3) The approach based on second-order context can obtain a lot of information in context and less noise will be produced.