Modeling User Interest Based On Search History

As information technology continues to evolve, the need for personalized services is growing faster. Currently, search engine systems usually provide search results by calculating the similarity of input queries and crawled documents, during which the users'personal background and interest are not taken into account. Thus different users get the same query result as long as they submit the same query, even if they have very different interests and backgrounds. This makes the traditional information retrieval techniques hard to meet individual needs. Facing massive network information, how to seize the user's personal interests, to provide users with personalized Web services technology has become an important research topic.The creation of user interest model is a crucial step in personalized services. The quality of the user interest model effect directly the accuracy and effectiveness of the personalized service delivery. The goal of modeling user interest is to dig infonnation of user's interest from historical data and then use the appropriate model represent.In this paper, a user interest modeling method based on user search history is brought. A user interest model is calculated by extract information from user's search logs and click documentation. Chinese word segmentation, document vector model and the text clustering and other methods are used in the process. The user interest model is constituted by a recent search word vector, a history search keywords vector, a document center vector and a query catalog vector. Query enrichment and search result ranking methods based on user interest model is discussed.