A Personalized Metasearch Engine Based on MultiAgent System

By providing unified access to multiple underlying search engines, metasearch engine is an intuitiveway to increase the coverage of the WWW. Great progress has been made in this area, butthe previous studies ignore the perspectives of users. This paper proposes a personalization mechanism for metasearch engine based on multi-agent system to improve precision ratio. The proposed mechanism obtains user interests from click-through data, schedules the appropriate underlying search engines according to the expertness model, and merges results based on user interest distribution. Moreover, it also has the ability to provide personalized result recommendation. Compared with the baseline results, experimental results show that the proposed personalization mechanism performs better on precision. The proposed metasearch engine is feasible for providing useful search results more effectively.