Speech Interaction Theory

Abstract

With the spreading of interactive speech system technologies, a clear need arises for theory which may adequately support the development of increasingly sophisticated but still restricted interactive speech systems. A complete and applied theory of spoken human-machine interaction would rigorously support efficient interactive speech system development from initial requirements capture through to the test and maintenance phases. It would include support for interaction model development and implementation, appropriate functionality design, usability optimization, interactive speech system evaluation and maintenance. Above all, such a theory would have to be based on the fact that the interaction models of today’s interactive speech systems are all task-oriented, they enable the system to carry out spoken interaction with users in limited application domains (Smith and Hipp, 1994). When combined with a basic level of meta-communication, or communication about the interaction itself, task-orientation is what enables current systems to successfully undertake spoken dialogue with humans despite their many limitations compared to human interlocutors. These comparative limitations may be briefly illustrated by taking a look at spoken human-human communication.