Background and Related Research

This book presents novel techniques for modeling spoken human-machine interaction to foster adaptivity in SDS. Primarily we are interested in the detection of critical dialog situations whose symptoms are certain emotions as well as particular observations made during an interaction. Both, emotions and these interaction observations are “patterns” that may be statistically modeled. Therefore, pattern recognition and machine learning approaches are pivotal to the following chapters. The presented approaches consequently do not rely on static rules, but may be derived and learned from data and may thus be easily ported to systems operating in new domains.