Representative And Ranking Of Text Database

Distributed text search is a search method which is highly effective in both range and speed. Meta-search is one of the important distributed text search method. Meta-search engine is a search engine based on many other search engines. It maintains the information of its component search engines rather than owns a database or index as other search engine. When a query comes, meta-search engine does not handle the query itself. Instead, it transfers the format of the query according to format of different component search engines and then dispatches the query to its component search engines. When the results are returned from component search engine, meta-search engine will merge all the results together and display the integrated result to the user.There are too many search engines now, but generally information the users want just distribute in some of them. If meta-search engine dispatches queries to all of component search engines, it will cause the delay of response. So we need to dispatch the queries to component search engines selectively. How to select database will have impact on quantity of search result. Representative and ranking of text database will help meta-search engine select the databases which are more likely to have the information users want. In this thesis, we are focusing on research in this area.In this thesis, we first discuss the significance of database selection in meta-search engine. Then we analyze the Query-Based Sampling and Hierarchical Database Sampling and then we focus on Adjusted Maximum Normalized Weigh which is used for database ranking in meta-search engine. We further discuss the defects and the improvements of adjusted maximum normalized weigh which include: adjustment of stopword list, adjusted maximum normalized weigh with feedbacks, adding synonym and correlation terms, modified term combination algorithm. Finally, we show the work that the improvement makes adjusted maximum normalized weigh more effective.