Research On The Aggregation Sorting Technology In Meta Search Engine

How to find the information what we need on the Internet has become the mainproblem. Meta search engine is based on component search engines. It merges theresults lists returned from multiple search engines to improve coverage ratio, but ithas also brought new problems. The results returned from multiple search engines aretoo large and many results are not relevant to the users’ queries. This directlyaffects the quality of the retrieval and greatly increases the cost of users’ queries. Inorder to help the user avoid the interference from useless information, we do researchon the technology of the user preference and web page. The dissertation develops auser page model based on web page classification and user interest classification. Onthe basis of this model we improve the sorting algorithm of mete search engine. Themain work and research achievements are as follows:After analyzing the relationship between users and web pages on search engine, amany-to-many model is put forward to match the relation in this dissertation. Byaccomplishing the conversion rules about interest and category, we build a basic userpage interactive model.Researching the algorithms of user interest classification, we propose a methodof getting user information based on the explicit feedback and implicit feedback. Thealgorithm of user interest auto update is also proposed, which is based on the UPImodel.We study the web pages automatic classification technology and complete thebasic set of automatic information classifier. On the base of UPI model web pagescategories update algorithm is proposed.We analyze two result sorting algorithms: web ranking order basedand association degree based. UPI-based sorting algorithm is proposed on thebasis of these two algorithms. This algorithm calculates the final score by consideringlocation score, user preference and matching between user and page. The value ofmatching can be calculated by UPI model.An experiment system based on the algorithms described above is implemented.The experiments show that the algorithms are valid and effective.