Learning User Profiles in NAUTILUS

Marco Gori, Marco Maggini, Enrico Martinelli, Franco Scarselli

Abstract: NAUTILUS is a Web recommender system that exploits a new approach to learn user profiles. The novelty consists of using a structured representation of HTML documents that allows us to split the page into logical contexts (lists, headers, paragraphs,...). The learning algorithm is based on a new neural computational model particularly suited to process structured objects.