AWARD NUMBER: 0546443

INSTITUTION: University of Minnesota, Twin Cities

NSF PROGRAM: Information and Knowledge Management

PRINCIPAL INVESTIGATOR: Gediminas Adomavicius

PROJECT TITLE: CAREER: Next Generation Personalization Technologies

PROJECT SUMMARY: Due to the rapid growth of the Internet and the continuously increasing accessibility to communication technologies and to a vast amount of information on the Web, the problem of information overload has become increasingly more visible in recent years. Various personalization technologies have been proposed to address this problem, including recommender systems that represent the most researched and developed personalization techniques applicable to various types of personalized offerings. Most current-generation recommender systems focus on recommending items to users and represent user preferences for an item with a single rating. Judging by the amount of attention this research area received in the last decade, this approach has worked well in certain applications. However, the inherent limitations of the recommendation framework are preventing recommender systems from tackling more complex personalization applications. In such applications, representing potentially complex user preferences with a single rating and ignoring the contextual information of personalization process is likely to negatively affect the predictive performance of recommender systems. We propose to develop an enhanced recommendation framework by introducing the ideas of context awareness, multi-criteria ratings, rating aggregation, recommendation flexibility, and non-intrusiveness into the recommendation process. Based on these ideas, the newly developed techniques will provide better personalization quality.

STUDENTS:

• Jesse Bockstedt (Ph.D. program, Information and Decision Sciences -- soon to be a faculty member at George Mason University)
• Sreeharsha Kamireddy (M.S. program, Computer Science)
• Praveen Kannan (M.S. program, Computer Science)
• YoungOk Kwon (Ph.D. program, Information and Decision Sciences)

Project 1: Multi-criteria recommender systems

Publications:

• G. Adomavicius and Y. Kwon. "New Recommendation Techniques for Multi-Criteria Rating Systems." IEEE Intelligent Systems. To appear. Working paper version can be found here.

Project 2: Recommendation query language

Publications:

• G. Adomavicius, A. Tuzhilin, and R. Zheng. "REQUEST: A Query Language for Customizing Recommendations." Under review. Working paper version can be found here.

Project 3: Variance-based filtering

Publications:

• G. Adomavicius, S. Kamireddy, and Y. Kwon. "Towards More Confident Recommendations: Improving Recommender Systems Using Filtering Approach Based on Rating Variance." Proceedings of the 17th Workshop on Information Technology and Systems (WITS’07), Montreal, Canada, December 2007. Working paper version will be posted soon.

Project 4: Visual mapping for trend identification

Publications:

• G. Adomavicius and J. Bockstedt. "C-TREND: Temporal Cluster Graphs for Identifying and Visualizing Trends in Multi-Attribute Transactional Data." IEEE Transactions on Knowledge and Data Engineering. Forthcoming. Working paper version can be found here.
• Preliminary version of this paper has appeared in: J. Bockstedt and G. Adomavicius. "A Visual Mapping Approach for Trend Identification in Multi-Attribute Data." Proceedings of the 17th Workshop on Information Technology and Systems (WITS’07), Montreal, Canada, December 2007.

Last modified: May 22, 2008