Dialog systems for automotive environments

The Center for Advanced Vehicular Systems (CAVS), located at Mississippi State University (MSU), is collaborating with regional automotive manufacturers such as Nissan, to advance telematics research. This paper describes work resulting from a research initiative to investigate the use of dialog systems in automotive environments, which includes in-vehicle driver as well as automotive manufacturing environments. We present recent results of an effort to develop an in-vehicle dialog prototype, preliminary to building a dialog system to assist in workforce training in automotive manufacturing. The overall system design is presented with focus on development of the semantic information needed by the natural language and dialog management modules. We describe data collection and analysis through which the information was derived. Through this process we reduced the parsing error rate by over 20% and system understanding errors to 3%.