Study And Design Of Personalized Meta Search Engine

With the development of web technology, Internet has been a global information space. The users often can not find the information they need in this situation. The advent of traditional search engine has solved the problem of the difficulties in web information retrieval in some degree. However, the overlapping of search results for the same query is no more than 34 percent, because different search engines have different retrieval algorithms and search scope. If users want to get relatively full scale, accurate search results, multi-search engines should be used. In this background Meta Search Engine appeared.Meta Search Engine is a type of search tool which implements its search functions by invoking other independent search engines. Generally, a Meta Search Engine has no its own database, which implements the information retrieval by integrating, invoking, controlling and optimizing the results of other search engines. Through a uniform interface, Meta Search Engine help users select appropriate independent search engines, then arrange the search results, finally return the results to users. Although it is convenient for users to retrieval in some degree, Meta Search Engine, as the traditional search engines, can not offer users the search results according to users' personalized requirement.In order to solve these problems, the thesis presented a personalized Meta Search Engine based on Agent technology, S4U. The system focuses on personalized service, which can choose the documents that users require from many documents. The system builds user profiles using Vector Space Model; modifies user profiles through learning the users' interests by agent to reflect user's interests in time. The thesis also presented a formalized description of member search engines, implements a general algorithm of extracting search results from member search engines. Thus the system solved the problem of customizing the algorithm of extracting search results from different search engines, and the problem that user can not dynamically modify source search engine list.The thesis presented the scheme of implementing personalized Meta Search Engine in the end. We also tested and analyzed the system, and the experimental results approved that the system had a good performance in information recommendation, precision and recall.