Semantic Web Recommender systems Cai-Nicolas Ziegler Institut fur Informatik, Group DBIs, Universitat Freiburg, Germany ziegler@informatik. uni-freiburg de Abstract. Research on recommendersystems has primarily addressed centralized scenarios and largely ignored open, decentralized systems where remote informa- tion distribution prevails. The absence of superordinate authorities having full access and control introduces some serious issues requiring novel approaches and methods. Hence, our primary objective targets the successful deployment and inte gration of recommender system facilities for Semantic Web applications, making use of novel technologies and concepts and incorporating them into one coherent 1 Introduction Automated recommender systems intend to provide people with recommendations of products they might appreciate, taking into account their past product ratings profile and history of purchase or interest. Most successful systems apply so-called social filtering techniques [9], dubbed collaborative filtering [6]. These systems identify similar users and make recommendations based upon products people utterly fancy Unfortunately, common collaborative filtering methods fail when transplanted into decentralized scenarios. Analyzing the issues specific to these domains, we believe that two novel approaches may alleviate the prevailing problems, namely trust networks, along with trust propagation mechanisms, and taxonomy-driven profile generation and filtering. One aspect of our work hence addresses the conception of suitable components specifically tailored to suit our decentralized setting, while another regards the seamless integration of these latter building bricks into one single, unified framework. Empirical analysis and performance evaluations are conducted at all stages 2 Research issues Deploying recommender systems into the Semantic Web implies diverse, multi-faceted issues, some of them being inherent to decentralized systems in general, others being cific. Hereby, our devised Semantic Web recommender system performs all recom- mendation computations locally for one given user. Its principal difference from generic, centralized approaches refers to information storage, supposing all user and rating data distributed throughout the Semantic Web. Hence its decentralized nature. We thus come to identify several research issues: et al. (Eds ) EDBT 2004 Workshops, LNCS 3268, pp. 78-89, 2004Semantic Web Recommender Systems Cai-Nicolas Ziegler Institut f¨ur Informatik, Group DBIS, Universit¨at Freiburg, Germany cziegler@informatik.uni-freiburg.de Abstract. Research on recommender systems has primarily addressed centralized scenarios and largely ignored open, decentralized systems where remote informa￾tion distribution prevails. The absence of superordinate authorities having full access and control introduces some serious issues requiring novel approaches and methods. Hence, our primary objective targets the successful deployment and inte￾gration of recommender system facilities for Semantic Web applications, making use of novel technologies and concepts and incorporating them into one coherent framework. 1 Introduction Automated recommender systems intend to provide people with recommendations of products they might appreciate, taking into account their past product ratings profile and history of purchase or interest. Most successful systems apply so-called social filtering techniques [9], dubbed collaborative filtering [6]. These systems identify similar users and make recommendations based upon products people utterly fancy. Unfortunately, common collaborative filtering methods fail when transplanted into decentralized scenarios. Analyzing the issues specific to these domains, we believe that two novel approaches may alleviate the prevailing problems, namely trust networks, along with trust propagation mechanisms, and taxonomy-driven profile generation and filtering. One aspect of our work hence addresses the conception of suitable components, specifically tailored to suit our decentralized setting, while another regards the seamless integration of these latter building bricks into one single, unified framework. Empirical analysis and performance evaluations are conducted at all stages. 2 Research Issues Deploying recommender systems into the Semantic Web implies diverse, multi-faceted issues, some of them being inherent to decentralized systems in general, others being specific. Hereby, our devised Semantic Web recommender system performs all recom￾mendation computationslocally for one given user. Its principal difference from generic, centralized approaches refers to information storage, supposing all user and rating data distributed throughout the Semantic Web. Hence its decentralized nature. We thus come to identify several research issues: W. Lindner et al. (Eds.): EDBT 2004 Workshops, LNCS 3268, pp. 78–89, 2004.
c Springer-Verlag Berlin Heidelberg 2004