Abstract
innsbruck.mobile is a mobile tourist guide for visitors of the city of Innsbruck. The system allows the retrieval of detailed information about sights, accommodation, events, dining out and weather forecasts as well as pushes recommendations from these categories to registered visitors via a short-messaging service. Personalising the information access and filtering of content according to the assumed relevance to its users was a key requirement for the system - especially in the context of mobile services with limited display capabilities. However, most real users conduct few and very short interactions with the system and therefore provide only sparse relevance feedback. As a consequence collaborative recommendation methods need to utilise multiple sources of user feedback to increase the accuracy on sparse rating data. We give a description of our system and conduct an evaluation based on this sparse usage data that includes more than 800 distinct user sessions.