Personalized Ranking Algorithm Based on User Interest Modeling

Abstract

Abstract. It's convenient for Internet users to access to web resources with a search engine. However, most traditional search engines can't provide personalized search results for users. In order to overcome this limit, we adopt a combination method of the explicit and implicit user modeling to build and update a user interest model. To be specific, we first build the user interest model with a topic-based representation method by the information offered by users. Then we update the model by considering the time factor and the user browsing behavior. As nouns are obviously more distinctive than other words, we give a greater weight for them. Based on this, we have improved the BM25 algorithm. Finally, a personalized ranking algorithm combining topic ranking and BM25 ranking is proposed. The experiments show that the personalized ranking algorithm based on user interest modeling can provide personalized search service for users.