Situated dialogue coordination for spoken dialogue systems

In this paper, we present a general framework and architecture for maintaining dialogue coordination in spoken dialogue systems, in which intended behaviors and goals are incrementally performed during the course of maintaining dialogue coordination. The dialogue structure emerges as a result from interaction between user and the dialogue system. The key feature of this design for the systems is to use multiple situated-agents for coordinating communicative acts that are realized as a hierarchy of autonomous behaviors by using a subsumption architecture. In this architecture it should be noted that the lower-level behaviors act autonomously for maintaining the dialogue coordination and are linked to the specifications from higher-level behaviors for dialogue management. In order to make the behavior of the system social, in general, the maintaining of dialogue coordination takes priority over the realization of intended goals of the system as a dialogue participant. We introduce an under-specification strategy for controlling the preference of the concurrent behaviors. This is in contrast to the classical, top-down approach to dialogue coordination.