Research And Implementation On Users' Interest-oriented Web Search Strategies

With the booming of Internet, information on the WWW has grown to a large quantity which has already gone beyond people's capability of dealing. It has become a huge obstacle that information is hard to find and retrieve and has made the usage of information on the Internet difficult. As a convenient tool of retrieving information on the Web, search engines emerge for solving the problem. At the same time, more and more attention has been paid to search engines, especially their performance. For most of the current search engines, the difference of their returned results only relies on the different keywords, i.e. for the same keywords used to search, the same results will be returned. In fact, different users may have different search purposes even if they use the same keywords. For this reason, the search engines are urgently required, which can provide the results accurately relating with users' search purposes and personalized services for users.In this thesis, the characteristics of current general search engine and the personalized search engine are analyzed. Based on the analysis, an idea of creating user model based on user search history and building users-oriented search engine based on the users'interest model is proposed. In detail, users'interest model is constructed on ground of users'previous clicks and the users'behavior is traced to update their interest model.To overcome the disadvantages of current ranking algorithms, User-Oriented PageRank (UOPR) is proposed on the based of users'interest model. UOPR incorporates the semantic distance between Web pages and topics, users' interest vector and some other factors into the re-ranking expressions, thus provides personalized services for users rather than queries.The experimental results show that the constructing and updating algorithms of users' interest model can represent users'preference well, the first-page re-ranking algorithm makes the result-list more close to users'preference, and the successive-page re-ranking algorithm implements the alternating search environment, which can create the successive page according to users'behaviors on-line.