A Framework for Personalizing Web Search with Concept-Based User Profiles

摘要

Personalized search is an important means to improve the performance of a search engine. In this article, we propose a framework that supports mining a user's conceptual preferences from users' clickthrough data resulting from Web search. The discovered preferences are utilized to adapt a search engine's ranking function. In this framework, an extended set of conceptual preferences was derived for a user based on the concepts extracted from the search results and the clickthrough data. Then, a concept-based user profile (CUP) representing the user profile as a concept ontology tree is generated. Finally, the CUP is input to a support vector machine (SVM) to learn a concept preference vector for adapting a personalized ranking function that reranks the search results. In order to achieve more flexible personalization, the framework allows a user to control the amount of specific CUP ontology information to be exposed to the personalized search engine. We study various parameters, such as conceptual relationships and concept features, arising from CUP that affect the ranking quality. Experiments confirm that our approach is able to significantly improve the retrieval effectiveness for the user. Further, our proposed control parameters of CUP information can adjust the exposed user information more smoothly and maintain better ranking quality than the existing methods.