Evaluating a User-Model Based Personalisation Architecture for Digital News Services

An architecture that provides personalised filtering and dissemination of news items is presented. It is based on user profiles and it provides mechanisms that allow the user to control and tailor to his own needs the interaction between three different sources of relevance judgements: the existing newspaper categorisation by sections, basic information retrieval on user selected keywords, and an additional operation of automatic categorisation against an alternative hierarchy of categories. These three tiers cover some of the most promising access methods for digital libraries. The proposed architecture has been implemented and evaluation results are presented, covering user response, system efficiency, and user preferences regarding the set of methods made available to them.