Research On Technology Of Personalized Information Retrieval Based On Multi-Agent

With the explosion of information accessible via the Internet, how to get useful information rapidly in huge-scaled web data is becoming more and more important. Current generic search engines satisfy users' search need in some extent. However, they completely ignore the difference of users' interests or background. Personalized Information Retrieval techniques to solve the problem emerge, which can provide with personalized search results.This paper discussed several important problem of the Personalized Information Retrieval techniques, such as the acquisition and description of user-interest, user-interest modeling method and the integrating method of the user-interest model and the retrieval model.In this paper, current information retrieval models are overviewed and analyzed firstly. Then a term co-occurrence based user-interest modeling method is presented. After that, to adapt to users'personalized behaviors, a cache-based model updating method is proposed. Next, as an integrating method of interest model and information retrieval systems, a personalized query expansion method is designed. And finally, the design, implementation and test of the personalized IR system based on multi-agent are described. This thesis focus mainly on three techniques: information retrieval modeling, user-interest modeling and query expansion method.Based on above techniques and researches, a personalized information retrieval based on multi-agent is established. The experiments show that the improvement of precision of Top 10 can promoted 16.25% and the improvement of mean reciprocal rank of top 10 results can achieved to 15.2%. Users can achieve better results that meet their current needs from this system. The experimental results also show that the system could also provide better personalized service for users with multi-interests.