Chinese Word Sense Disambiguation Based On Parsing Tree

The polysemy phenomenon of Chinese has brought many difficulties tonatural language processin (NLP). The most important thing in natural languageprocessing is to solve the problems of ambiguous words. Word sensedisambiguation (WSD) is used to make computer understand the accuratemeaning of an ambiguous word in a certain context. It is very important to thefield of natural language processing, such as information retrieval, machinetranslation (MT), automatic text classification, automatic abstract and many otherissues.This paper studies word sense disambiguation based on statistical learningmethod in machine learning. This method has good capacity to cope with thedevelopment of language. It is the popular method to tackle the problem of wordsense disambiguation. In this paper, the research contents are divided into thefollowing parts:Firstly, we study the various classification methods and characteristics ofword sense disambiguation. At the same time, we study the authoritativeevaluation systems of word sense disambiguation and put forward the possibleproblems in the process of word sense disambiguation and solutions.Secondly, we study feature engineering in word sense disambiguation,including feature extraction and feature selection. In feature extraction, weintroduce the feature extraction method based on the sliding window and theparsing tree. At the same time, we focus on explaining the establishment processof parsing tree and the algorithm of feature extraction method based on parsingtree. In feature selection, the word-based method to extract features is researched.We fuse the syntactic information and part-of-speech information to selectfeatures. At the same time, a semantic disambiguation classification model basedon parsing tree is optimized according to the robustness and applicability of simple bayesian model.Finally, we study the effect of word sense disambiguation on machinetranslation system. The module of word sense disambiguation is integrated intomachine translation system. Although this research is not perfect and the resultsare not satisfying, it can also provide some helps for machine translation.