Measurement, Analysis and Modeling

Abstract

Better understanding of conversation paves the way towards better conversational systems. In this chapter we shed light on the practical aspects of multi-modal interaction analysis towards a better understanding of conversation as a phenomenon. On the one hand, investigators need to take great care of methodological issues, since conversation involves plenty of subtleties. Incorporating physiological signals allows us to base our understanding on a more solid foundation than merely depending on audio-visual data, which is extrinsic from the viewpoint of mental processes. Collaborative support tools help annotators share their experiences thereby improving the efficiency and quality of the annotation process. On the other hand, investigators are encouraged to learn from past experiences in terms of how experiments were conducted to derive useful insights. For this reason we report in this chapter three case studies that showcase different issues in measurement, analysis and modeling of conversation and interaction in general.