Modeling user's preference in folksonomy for personalized search

Personalized search based on the users' preference has been extensively studied in the field of information retrieval. As a typical representative of web2.0, social tagging not only allows users to better describe and manage web resources, but also provides a great opportunity for the personalized search research, since it contains abundant public personal information. In this paper we propose a user preferences model based on tag clustering. This model is inspired by the Data Field theory. It can depict users' different aspects of interests and dynamically generate users' preferences against different search queries. We apply this model in the data set of flickr. The final ranked pictures are the combination of keywords and users' preference matching. The experiment proves that our method is better than both non-personalization method and common personalization method.