TOAST: A Topic-Oriented Tag-Based Recommender System Guandong Xu 1 , Yanhui Gu 2 , Yanchun Zhang 1 , Zhenglu Yang 3 , and Masaru Kitsuregawa 3 1 School of Engineering and Science, Victoria University, Australia 2 Dept. of Information and Communication Engineering, University of Tokyo, Japan 3 Institute of Industrial Science, University of Tokyo, Japan Abstract. Social Annotation Systems have emerged as a popular appli- cation with the advance of Web 2.0 technologies. Tags generated by users using arbitrary words to express their own opinions and perceptions on various resources provide a new intermediate dimension between users and resources, which deemed to convey the user preference information. Using clustering for topic extraction and incorporating it with the cap- ture of user preference and resource affiliation is becoming an effective practice in tag-based recommender systems. In this paper, we aim to address these challenges via a topic graph approach. We first propose a Topic Oriented Graph (TOG), which models the user preference and resource affiliation on various topics. Based on the graph, we devise a Topic-Oriented Tag-based Recommendation System (TOAST) by using the preference propagation on the graph. We conduct experiments on two real datasets to demonstrate that our approach outperforms other state-of-the-art algorithms. 1 Introduction Tag-based services, e.g., Del.icio.us 1 , Last.fm 2 , and Flickr 3 , have undergone tremendous growth in the past several years. All of the above services allow users to express their own opinions on resources with arbitrary words. A pri- mary concern in personalized recommender systems is to present users with instrumental means for navigating the resources that are most relevant to their real information needs. In social annotation systems, tags serve as an interme- diate metadata between users and resources conveying the users’ navigational preference, therefore the key challenge in social annotation recommender sys- tems is to accurately capture user preferences through tags and make use it for personalized recommendation. A typical social annotation system has three kinds of entities: user, resource and tag. The user prefers some resources and annotates with some words. The 1 http://del.icio.us/ 2 http://www.last.fm/ 3 http://flickr.com/ A. Bouguettaya, M. Hauswirth, and L. Liu (Eds.): WISE 2011, LNCS 6997, pp. 158–171, 2011. c  Springer-Verlag Berlin Heidelberg 2011