Asynchronous Dialogue Management: Two Case-Studies

Much of the human-machine dialogue research in the literature tacitly assumes a “synchronous” dialogue model; user talks, system acts, system replies. In particular, the user is not supposed to interrupt the system, neither when it talks nor acts. In this paper, we argue that the synchronous model is not appropriate for most interesting real-life applications, but there is a need for asynchronous dialogue models where the user has the possibility to interrupt the system at any time. By referring to two implemented asynchronous dialogue systems, we try to pinpoint what implementation requirements such a dialogue model entails, and we also outline some of the theoretical implications of asynchronous dialogue management.