| According to others' statistic, there are millions of new documents and files adding to Internet every day. Now the sum of web pages is rising sharply, and reaches to several billion. It has been a critical problem on how to find necessary information in Internet. Search engine appeared under the situation. However, along With vast number of data coming into being in Internet, the best search engines only have no more than 30% coverage of global web pages. Moreover, due to different scopes and algorithms, the repetition of search results for the same query is no more than 34% for different search engines. In order to solving these problems, Meta search engines occur.Although the appearance ofmeta-search engine has being taken great conveniences and advantages, there are also being some important problems baffling the development of it, such as the limited net resource, the number of displaying results and search time and so on. In the text first we will introduce some knowledge about search engine, meta-search engine and optimal search theory, then we'll show how to apply optimal search theory and the way based on learning to build a optimal model to improve the performance of current meta-search engine model. In the new model, we'll allocate net resources, choose member search engines and configure the number of displaying results and search time in a optimal way, so users can get the optimal search results and performance under specified number of displaying results and search time, and embody a human way completely.In the end, we designed and implemented a parallel meta-search engine system based on some source code in Internet. In relation to other meta-search engines, this system can be used in a human way. And we analyzed the query time and the precision ration. Comparing to common search engines and original meta-search engines, we proved that this system is better in precision ration and is also better than original meta-search engines in responding time. |
PDF Full Text Request