Adaptive personalization of multimodal vehicular interfaces using a hybrid recommendation approach with fuzzy preferences

User-adaptive and situation-aware presentation of information on multimodal interfaces is an important research area for coping with the increasing information load on the driver. This paper describes a comprehensive recommendation approach for inferring vague individual preferences under uncertain conditions by using fuzzy preference relations. The approach was applied to rank internet information on a multimodal driver interface in the vehicle. New information gets successively downloaded to the car by using digital media broadcast and the driver can query information (e.g. press review, cheap petrol stations, snow reports) by using natural language queries. Incoming information elements are gradually matched to ontology categories by a fuzzy membership value based on term frequency. User perferences are learned from interaction and recommendations for content items is based on the resulting preference model. Three different types of fuzzy relations are aggregated to an overall preference score: explicit preferences, implicit preferences, and global preferences. The later influence preference values of one driver based on other drivers in vehicular ad-hoc networks. Therefore, the approach is not solely based on the presence of one kind of interaction data. Furthermore, the situation is modeled using the concept of granules and influences the overall preference score.