Personalized Retrieval in Social Bookmarking

Abstract

Users of social bookmarking systems take advantage of pivot browsing, an interaction technique allowing them to easily refine lists of bookmarks through the selection of filter terms. However, social bookmarking systems use one-size-fits-all ranking metrics to order refined lists. These generic rankings ignore past user interactions that may be useful in determining the relevance of bookmarks. In this work we describe a personalized ordering algorithm that leverages the fact that refinding, rather than discovery (finding a bookmark for the first time), makes up the majority of bookmark accesses. The algorithm examines user-access histories and promotes bookmarks that a user has previously visited. We investigate the potential of our algorithm using interaction logs from an enterprise social bookmarking system, the results show that our personalized algorithm would lead to improved bookmark rankings.