Research On Implicitly Interactive Personalized Search Engine

With the development and popularization of Internet in the world, more and more information has been published on the Internet. Users of Internet face the problem of information overloading and can not easily find information available with much abundant information on the Internet, so personalized search engine become an effective method to improve the situation or resolve the problem. Personalized search engine and personalized technology become hot research topics, and some research achievements come to existence.The thesis presents the status of the personalized search engine and existing key technologies, as well as the deficiencies of personalized search engines. By analyzing the features of user action in the course of user using search engine, a new research framework of personalized search engine called Implicitly Interactive Search Engine is proposed which based on implicit information collection of user action and implicit feedback technology. The research of the framework is meaningful in both theory and application. The main works in the thesis are listed as follows:Firstly, the thesis presents basic theory of Implicitly Interactive Search Engine, and the three-layer architecture and its communication mechanism are proposed.Secondly, the thesis presents the implicit user information collection method which based on the architecture of interactive search engine; and the collection contents and storing format is also presented. On the basis of information collected from users, the thesis describes the methods and steps of creating user profile based on which the method of user modeling and query expansion are proposed.Thirdly, on the basis of the user instant feedback collected through implicit message, a method of user model updating based on implicit user feedback technology is proposed. By analyzing user behavior and characteristics of the ranking features of search engine, a personalized ranking method based on cache is proposed. According to the user behavior and the characteristics of personalized search engine, User Action Index and User Action Precision as evaluation indicators of personalized search engine are proposed.Finally, an experimental system called ISEP proves the feasibility of three-layer architecture of Implicitly Interactive Search Engine. The experiment shows that query expansion method based on user interests can improve the precision of member search engine to some extend. In addition, average User Action Index and average User Action Precision demonstrate that ranking method based on user cache has certain advantages. It also shows two evaluation indicators in the performance evaluation of personalized search engine are consistency.