Personal Collaborative Filtering for the Internet: The GroupLens Research Project

ABSTRACT
Since 1992, the GroupLens research project has been exploring the use of collaborative filtering to ease information overload. After building several prototype systems, conducting a public trial filtering Usenet News, and starting another public trial filtering and recommending movies, we have learned a great deal both about collaborative filtering and about addressing information overload in a dynamic environment. This talk introduces the problem of information overload and discusses several approaches to alleviating overload. It then discusses collaborative filtering in detail, including a discussion of collaborative filtering algorithms and empirical results from our experiements. The talk concludes with an analysis of the successes and failures of collaborative filtering along with some reflections on the process of developing and marketing a commercial version of the system--the GroupLens Recommendation Engine.