User-Adapted Car Navigation System using a Bayesian Network -Personalized Recommendation of Content

Recent car navigation systems now provide more content than ever. However, retrieving and selecting such content poses safety issues to users, especially drivers. Moreover, usability issues arise from simple user interfaces. Thus, it is important for the system to recommend content adapted to the user's preferences and situations automatically. In this paper, we analyze the validity of applying a Bayesian network to a user preference model of a content recommendation system in cars. We also present a practical way of building models using an information criterion as well as domain knowledge and an incremental learning method to adapt to individual users.