Research Of Chinese Word Sense Disambiguation Based On Hownet

Word sense disambiguation has always been a hot research issue of naturallanguage processing. As an indispensable link of natural language processing, it hasbeen widely used in many fields, such as machine translation, information retrieval,text analysis, automatic summarization, knowledge mining, etc.This paper first introduces the research background, significance, history andpresent situation of word sense disambiguation. It also analyzes and compares thepresent methods of disambiguation. On these basis, it proposes a method of Chineseword sense disambiguation based on Hownet, which combing the useful knowledgein Hownet, extracting the correlative words in the context by using improveddependency grammar analysis, determining the correct meaning of the ambiguousword through computing the value of the semantic relevancy.The highlight of this method is shown as below:1. It is based on Hownet. Hownet is the knowledge resource and semantictagging system. The information in Hownet can be used to direct disambiguation.2. It proposes the context choice based on improved dependency grammaranalysis, which takes relations of dependency, term relevancy and semantic similarityinto account, to extract the correlative words of ambiguous word. It overcomes thedisadvantage of “context window” and makes the disambiguation system moreeffective. And its accuracy has been improved by27.2%compared to the formermethod.3. It proposes a disambiguation method based on semantic relevancy, computingthe value of the semantic relevancy by taking term similarity, term relevancy andexample factor into account, and finally finishing the disambiguation. Compared withdisambiguation based on similarity, our disambiguation accuracy has been improvedby11.8%and the accuracy has been improved by9.1%while compared todisambiguation based on relevancy. The experimental result shows that the method in this paper works well.