Enhancing Search Engine Performance using Expert Systems

Search engines of today do a great job of sifting through billions of pages of Internet content and returning search results highly relevant to user queries. However, in localized implementations (a local university search or an Intranet search of a private company), the same search engine technology usually has less than satisfactory performance. The technology that works well on billions of pages of general content doesn't work well on a much smaller scale of closely related content. In this paper, we analyze the performance problem in localized search engine implementations and identify specific performance issues through examining search logs. Our proposed solutions to those issues are based on utilizing an expert system where the fixes to the search issues are defined as a set of rules. We conduct experiments with California State University, Sacramento Web site, and the preliminary results indicate that when applying those rules to search engine queries and search results, search engine performance and user satisfaction are improved