Messaging is one of the top applications requested by customers for in-car usage. Our goal is to make it as safe as possible while expanding features and comfort by adding intelligence. One important ingredient for this is to know about the content and topic of the message. This thesis deals with the challenge of how to extract specific information from emails that can be used for this purpose.

"I am stuck here in traffic at Kreuz Stuttgart and its incredibly hot outside". This tweet consists of three contextual parameters (traffic situation, location, outside temperature). We are interested in how messages like this can be automatically generated and suggested to the driver. An additional question is in how far this form of communication may decrease frustration / aggression.

The goal of the work is to extend an existing driving simulation solution developed at DFKI. Extensions may be either on the computer vision side (incorporating new models, scenes) or in the experimental side (new functions, new task, interfaces to other system components).

SPEARCONS (speech earcons) are compressed sounds that are produced by speeding up spoken phrases, even to the point where the resulting sound is no longer comprehensible as a particular word (Walker et al., 2006). These unique sounds are analogous to fingerprints because of their acoustic relationship with the original speech phrases.

Today, SPEARCONS are created by converting the text of a menu item to speech via TTS and speeding it up using a pitch-constant compression algorithm. Typically, SPEARCONS are prepended to the spoken menu item, which allows faster learning and navigation of the auditory menu (Myounghoon et al., 2009). Walker et al. (2006) showed that SPEARCONS resulted in faster and more accurate performance than other auditory cues for a search task.

- observing the driver eye gazes to learn when s/he looks at the navigation display in order to reassure

- the "do something" button: How can we predict what the user wants when we know that s/he wants something ?

- What kind of graphical abstractions are usefully / can / should be done under which circumstances?
collaboration with SIM-TD Partners who develop the underlying (data) components.

We have a number of topics available related to ongoing / planned user studies. 1) Using eye tracker for explicit interaction in the car (e.g. "mouse-over"). 2) Evaluation of a new driving simulator software (comparison with ISO standardized Lange Change Task). 3) Referring to outside objects with pointing gestures (iPhone). 4) Multimodal interaction: combination of speech and tangible interfaces (follow-up study).