A Framework for Guiding the Museum Tours Personalization

The importance of web access personalization to cultural heritage has been recognized by many museums. Recent research in recommender systems, information retrieval and data mining has facilitated the development of personalized applications. This research provides a number of intelligent technologies for supporting user navigation, information filtering, and other important processes of a user-centered interactive information exchange between museum websites and their visitors. In this paper we study some of the challenges of personalizing museum tours. We focus on (i) the efficient learning of user models in offline and online settings, and (ii) the scientific evaluation of a personalization effect. Given few examples of the preferences of a user, the system must learn to suggest the most relevant artworks. This setting is particularly important for the development of an interactive museum tour. An additional challenge here is to learn user preferences quickly to be able to start recommending relevant artworks as soon as possible. We review the basic approaches that have been (or potentially can be) used for personalizing the access to the cultural heritage. We also present a formal framework for evaluating online and off-line learning user models. In the off-line setting, the user provides the system with some labeled examples from which the system must generalize. In the online setting, however, the system can interactively query the user's interest. Then, we discuss the challenges of scientific evaluation of personalization techniques and personalization process. Finally, we present a methodological framework for guiding a museum tour personalization and discuss the potential utility of this framework.