Adaptive and multimodal interaction in the vehicle

In this paper, a system architecture for personalized and user-adaptive information presentation is described. In order to reduce cognitive load, interaction is done at a multimodal interface using visual information and speech as output channels, and manual input and speech as input channels. Most recent information from the Internet can be queried by natural language question-answering dialogs using frame-based dialog management. A fuzzy recommendation approach is described in detail and used in the system for structuring content items according to the user preferences. In addition to such a personalized visual list of topic areas for an initial choice, an adaptive ranking is applied to acoustical speech output. The item that relates best to the driver´s preferences is read out first by text-to-speech. Therefore, the first items presented to the user are ought to be of high interest. Cognitive load can potentially be further decreased by our approach for the task of choosing one item out of a sequence of alternatives in working memory. The choice can be immediately performed by using. the barge-in feature.