Document Refinement Based Query Expansion

The rapid progress in Internet technology provides us plenty of message resources, but also makes the information searching to be difficult. Information Retrieval System is used to resolve this problem. But there are too many kinds of jamming, for example, the synonyms, the inaccurate information put in the interactive interface by users,etc. which makes the recall and precision of searching low. To improve the performance of information retrieval system, the Query Expansion was presented in the 1970 s. The emergence of this technology has caused the attention of the researchers, which has some research meaning and practical value.This dissertation mainly studied the following parts:Firstly, the dissertation summarizes a study summarization on information retrieval technology both at home and abroad, the research background and significance of query expansion technology. Analyze the fundamental principle, the application situation and the retrieval performance of Vector Space model and query expansion technology. The query expansion technology has, query expansion based on global document analysis, local analysis, user query log, semantic dictionary, etc.Secondly, this dissertation proposes Document Refinement based query expansion. This dissertation chooses the LCA as the research object, improving the defect that depending on the initial retrieval document. The thought of document refinement based query expansion is, First of all, combine the document refinement with semantic dictionary, which expands the query words, to improve the correlation of the first result document sets. Combine the document refinement thought with the semantic dictionary to limit the layer of query expansion, calculate the similarity of two concepts of WordNet to conclude the semantic expansion words, which improve the recall and precision of retrieval system. Organize and sort the first n documents in the last document sets to improve the value of P@N and meet the retrieval demand of the user.Finally, integrate the query expansion module with open-source frameworkNutch. Calculate the evaluation indices of sets by searching the test sets, Contrast this system with traditional retrieval system without query expansion module, the retrieval system with the query expansion based on local context analysis, which prove that the system has relative increase in retrieval performance.