The Research And Realization Of Several Key Techniques For Question-Answering System Based On Search Engines

The advent of search engine has caused a revolution in the ways of information retrieval on the Internet, and helped people much to find resources they wanted in the vast information space, this is also why search engine becomes one of the most important and popular Internet applications. However, as Internet information resources accumulates, traditional search engines gradually expose more and more shortcomings, such as miscellaneous search results, overlong search time, low accuracy, etc. And therefore, research on search engines with new-style, high efficiency, and humanization, or new applications based on search engines, is a profound issue.This paper analyzed and investigated two important research directions in the area of information retrieval Automatic Question-Answering System and Meta-Search Engine, and combining the advantages of both, this paper brought forward a design plan on a question-answering system based on search engines, and do research and some realization for several key techniques related. So far, we have accomplished much work, which is as follows: (1) Using the layered and modularized design methods as the system's design principle, we have defined the three-layer architecture of the system, that is, user/application layer, mediator layer and search sources layer, we also accomplished the design work of the system's architecture; (2) We researched the key techniques related to the system, and also prepared for the difficult points of realization theoretically and technically, such as put forward the transform approach for member search engines'query URLs, and the ordering method of search results with unified correlation measure, etc; (3)We have completed the realization work of search result processor, and made an evaluation of the system's performance.Compared to traditional search engines, the system researched in this paper can make a better understanding of users'search demands, and thus can discover answer positions in a more precise way, but search time of the system may be affected by reason that in the process of information retrieval, the system has to call more than one member search engines, this is some sacrifice for improving search accuracy. In the process of follow-up research and implementation, more energy will be put on improving the semantic understanding ability, perfecting the FAQ knowledge base's establishment and management, and also strengthening stability and flexibility of the system.