Analyzing cross-system user modeling on the social web

Abstract

In this article, we analyze tag-based user profiles, which result from social tagging activities in Social Web systems and particularly in Flickr, Twitter and Delicious. We investigate the characteristics of tag-based user profiles within these systems, examine to what extent tag-based profiles of individual users overlap between the systems and identify significant benefits of cross-system user modeling by means of aggregating the different profiles of a same user. We present a set of cross-system user modeling strategies and evaluate their performance in generating valuable profiles in the context of tag and resource recommendations in Flickr, Twitter and Delicious. Our evaluation shows that the cross-system user modeling strategies outperform other strategies significantly and have tremendous impact on the recommendation quality in cold-start settings where systems have sparse information about their users. © 2011 Springer-Verlag.