Generating Personalized Destination Suggestions for Automotive Navigation Systems under Uncertainty

Programming a car's navigation system manually takes time and is error-prone. When the address is not handy, a cumbersome search may start. Changing the destination while driving is even more problematic. Given a modern car's role as an information hub, we argue that an intelligent system could in many cases infer the right destination or have it among the top N suggestions. In this work, we propose a personalized navigation system that is built from three main ingredients: strong user models, knowledge source fusion, and reasoning under uncertainty. We focus on emails as one particular knowledge source, exploring the uncertainties involved when extracting empirical data of email appointments.