Studies Of Enhancing The Performance Of WEB Information Retrieval Algorithms

Web information is avalanching nowadays as human knowledge accumulates and network applications become popular. There is more information that can be people acquire, and more difficulty comes when they want to get the useful information. People can not make use of the information effectively without good searching facilities.Information retrieval (IR) technology can help people search in large amounts of text data. But WEB information is unlike common text on the display mode, and the WEB has its own topology structure. We studied how to use the character of WEB to improve the traditional information retrieval algorithms on the WEB Track task of TREC2003, and got some achievement on it. Since no solvent on the perception of nature language, the computer can only judge the relativity by whether the words in user's query appear in the document. The retrieval system can not work intelligently, which make it difficult for the users to construct their query. The pseudo relevance feedback method can help the retrieval system to rebuild user's query, for a more precise description of user's requirement. But the pseudo relevance feedback method is a statistical one, so that there is a problem how to optimize the parameters. We studied the factors that influence he feedback performance in order to find a stable measure to optimize the parameter, and bring forward a measure to adjust the parameters dynamic. We regard the query as expression of information, and characters can be chosen according as mutual information, which can get a new query having the minimal uncertainty. By experiment on the data set of TREC2002, it is proved that using mutual information in information retrieval can improve the retrieval performance availably.