Personalized Search Based On Social Tags

With the drive of internet technique, the usage of search engine becomes more and more mature. Kinds of search engine become the advantaged tool when search information on the web. But because users'demand of information personalization is more and more exigent, it requires search engine improve its search efficiency. The emergency of web2.0 and its relevant technique brings a new surprise to the whole internet. Among this the most representative technique is social annotation which brings new chance to improve the personalization service quality of search engine.The research on personalization of search engine focuses on how to get the accurate personalization character of the users and how to supply personalization service according to the acquired users'preference. This paper first generally analyzes the pivotal technique in traditional weblog mining and summarizes the main method of data acquisition, data filtration, model expression, model study and update. After we find the problem existed in traditional weblog mining, in this paper, combined with the method of traditional weblog mining and the disposal of social annotations, we gain the characteristic and provide personalized recommendation to users.Social annotation web supplies information which are forwardly annotated by users, compared with traditional weblog mining, studying the users'preference from this information resource must be more accurate and external. Besides, the annotated action of users itself deserves being mined. The frequency of annotated action for one kind of resource reflects the attention degree of this interested point. Users can always find high quality resource earlier, so according to this character, we can recommend this resource to other users who have the same interests.The major work in this paper include the following three parts:(1) Describe the users'historical search records and annotated data with matrix, give description to the key term in search records and annotated data. Finally we get the user character model according to this matrix.(2) The expression based on users'character discusses the existing main method of learning and updating user model, and combine with the advantage we propose a adaptive method to learn the user model. (3) After we get the users'character, aim at the problem of data sparseness about normal users, we propose a personalized recommendation method based on cooperating filtration, meanwhile, according to combination with the users'annotated action and normal users'character, we provide users with the service of personalized communion.