Information Retrieval And Related Research Based On Query Log

Internet offers us abundant information resource. However, the huge quantity of information raises the problem for Information Retrieval System. In this information sea how to find the information that users need fast and correctly is a hot question. Due to the different understanding to the same information object between web users and authors, they usually use inconsistent queries to define the content of information. Additionally under restrict of key phrase match technology, the performance of information retrieval mostly depends on the query submitted by user. So our paper proposes a method of query expansion based on the query log. This method could expand user's query through finding the similar queries on meaning and so it would help users find their required information more easily.Firstly we arrange a relationship analysis on the Query-Result graph generated by a search engine and get the nominative query set. Secondly a snippet analysis is used to check every nominative query whether it is similar with original query on meaning. A snippet is a segment of description for a result URL and it could be found in the result of search engine's query request. At this step, we use an automatic key phrase extraction method to get the key word of the snippet of original query. Through matching key phrase with each nominative query's snippet set, we could compute the relevance score of every nominative query and get the final relevant query list.We also demonstrate our method on a data set of hundreds of thousands of queries. The experiment proves that our method could visibly improve the performance of query expansion.In the finality, the problems requiring further studies are discussed.