Research And Implementation Of Multi-Agent Systems For Intelligent Meta-Search Engine

With the growing amount of information on the Internet, search engines have become a more and more important means for obtaining useful information and one of the tools essential to daily life. However, information retrieved from a single search engine is limited; at the same time, various types of search engines which are good at a specific field exist nowadays; and the fact that an increasingly types of vertical search engines are presented enlarge the difference among each search engine. As a result, it is often necessary to search the contents by shifting to various kinds of search engines; but this will reduce the efficiency of information retrieval. Meta-search engine, a search tool with a universal interface for user to search several search engines simultaneously and merge results from these search engines, can lighten users' burden and improve the efficiency and comprehensiveness of retrieval. The meta-search engines are developed in a trend to intelligent and personalize. By analyzing users' retrieval interest, the meta-search engine can ensure the comprehensiveness along with the accuracy of search results. The current researches on meta-search engine mainly focus on the selection of individual search engines, the result merging algorithm and a well-designed architecture for improving the performance of meta-search engine.The paper mainly discusses about the research and implementation of a multi-agent systems for meta-search engine. First of all, a situation analysis of the currently existing meta-search engines and the application of multi-agent systems in meta-search engines are made. And some related technologies is reviewed, including technologies related to meta-search engine, theories about Agent and the JADE development framework used in this paper. And then, the design of an agent framework model for meta-search engine and description of our design of our mechanism of traditional web-based applications and multi-agent systems are discussed, including the structure, role, ability and behavior in the lifecycle of each agent. After that, by a combination of the design and JADE development framework, an implementation of the proposed multi-agent systems for meta-search engine is presented. In additional, a method of result merging method is presented to enhance the personalization of our system is introduced.A test and discussion on the presented system shows that the designed agents can collaborate with each other to fulfill the target of meta-search engine. And also, a test on the personalization of the system show the presented result merging mechanism do has efficient in improving the quality of search result.