| The internet develops rapidly in various fields after it appears. Knowledge and information increases exponentially. Meanwhile, the excess of knowledge makes the Internet users much trouble. To solve the problem of information overload and knowledge explosion, many commercial search engines has become an important means of information retrieval. Through a search engine, users enter the relevant key words and will get the information contained key words. But the mismatch between documents and queries is one of the most important matters to influence the effect of retrieval. To solve this problem, we can use query expansion according to statistics or semantic recognition.The techniques of query expansion are as follows:Query expansion based on global analysis, query expansion based on local analysis, query expansion based on local context analysis. Query expansion based on semantic dictionary, etc. Global analysis and local analysis based on statistical data, and can not eliminate the semantic bias. Query expansion based on semantic knowledge dictionary can eliminate the bias and doesn't require the support of large-scale corpus, and the disadvantage is non-real time. The purposes of this paper is to use the advantages of statistics and semantic. This paper makes a series of studies on query expansion.The innovation of this paper lies in the following two aspects:1. It proposed a new algorithm to calculate the similarity between words. Based on this algorithm, it designs a query expansion model which combines with semantic concept tree and local analysis, and get the extended words by semantic concept tree and local context analysis. This paper also discusses multi-keywords, which makes this algorithm have more practical value.2. It designs an iterative algorithm to calculate the similarity between words and similarity between short texts, discusses the time complexity. Finally, by calculating the correlation coefficient, it proved to be closer to people's judgment on similarity. |
PDF Full Text Request