Query Expansion Based On Semantic Analysis And Local Documents

With the development of modern information technology and network technology，large amounts of information have been put on the web and increase rapidly. So how toaccurately and efficiently get the needed information from the web becomes an importantissue. Currently most information retrieval system use text matching technology toretrieve document，but there always be difference between user’s query and their potentialdemands. Therefore many documents which don’t contain user’s query but tally withuser’s demand can not be retrieved. At the same time, a lot of queries user submit are veryshort and users always just pay attention to the top retrieved documents. All of theseproblems have affected the performance of information retrieval system.A query expansion method based on semantic analysis and local document has beenpurposed to solve the problems above and improves the performance of informationretrieval system. This method uses relationship between concepts extracted from externalcorpus to analysis semantic similarity of concepts. When user submit a query, according tothe semantic similarity the method first gets three expanded word sets from WordNet,Wikipedia and local documents which retrieved by the original query and then combinesthe expanded word sets and allocates new weight to select good expansion words used toform an expanded query. Finally use vector space modal to rank documents retrieved bythe expanded query and return them to the user. On one hand, this expanded method takesadvantages of the local query expansion method and improves the quality of expandedwords. On the other hand, it has compensates the deficiency of the local expansion methodas well as the expansion method based on external corpus.A lot of experiments have been done on standard dataset of international text retrievalconference and the experimental results show that compared with original query and queryexpansion method just based on Wikipedia, this method can effectively improve theperformance of information retrieval system. With the increase of expanded word, theeffect of improvement is also obvious.