This paper presents a framework for improving recommender systems by exploiting the users tagging activity. The underlying principle is that the activities performed by users, specifically annotation-based activities, can be tracked and used by a recommender as a powerful feedback to enrich the model of the user it dynamically builds. The paper presents also a prototype, iCITY, developed to test the validity of the framework. iCITY is an adaptive, social, multi-device guide that provides suggestions about cultural resources and events. Its evaluation has been carried out at different stages of its development and covered several features: the accuracy of recommendation, the role of user tags in the definition of the user model and the usability of the adaptive user interface. The results are reported in the contribution and seem to be a confirmation of the validity of the approach.
展开▼
