Individualized WEB Searching Based User Interest

Today, major Search Engines fail to consider user interests and user differences. Search Engines only execute the user description passively and rigescently. As long as the same key words are keyed into the Search Engines, the same results will occur, and the hits are thousands upon thousands. Thus, the problem in the quality of searching results is serious. Therefore, it is a very important task to consider the user's interests in Search Engines.This article first studies the forms and changes of user interests, There are three characteristics of user interests: variability,stability and evolutivity. Applications of the three characteristics to user modeling and personalized recommendation are provided. In order to save the time of the user to browse the web sites, the outdated key words stored in user interest descriptions must be deleted regularly. This is effective to update user interests and remove the outdated interests to avoid interference. A default word frequency threshold is proposed in judgement of the user interests. Words with frequency lower than the default frequency are not valid .This filtrates the influence of the user's temporary interests.The importance of the Web page titles is studied. Web page title is used to represent the web page content. An arithmetic using the statistic method of the key word occurrence to form the initial user interest descriptions is provided. According to the given set of web titles, a user is inquired interactively to tell its interest in web titles. And then the arithmetic decomposes these titles to key word clusters that describe the user interests through Chinese Words Divided Syncopation Technology, and finally the initial user interest description file is formed.An intelligent local Agent is designed to trace four actions of the user browsing the web site. These four actions include: browsing times, staying time in a web page, page saving and searching key words. After catching these four actions, the web page titles are decomposed into key word clusters to update the user's interest description file. So, the user interest Description file is automatically updated.An arithmetic that judge user interests is provided. This arithmetic calculates according to the formed user interest description to realize individualized Web page recommendation.An experiment of user modeling and user interest tracking is executed and the formation and changes of user interests are simulated. Filtration of the influence of the user's temporary and outdated interests. User's long and steady interest becomes prominent. The experiment presents a favorable result.