Abstract

Adaptive Dialogue Management - How to make human machine dialogue more user friendly.

Adaptive dialogue management is a challenge in the field of spoken language dialogue systems. There are a lot of aspects that have to be taken into consideration in order to create adaptivity.

This presentation tries to discuss some of the aspects and to provide solution propositions for them not only limited to spoken dialogue but also regarding multimodal dialogue systems, as well.
Firstly, as human machine interaction can only be adaptive with respect to the user of the system, there will be a short discussion about how to evaluate different ways of system reactions. For this purpose the design of a WoZ-study will be described.
Secondly, adaptive dialogue management requires specific dialogue models that facilitate adaptivity with regard to different issues like change of initiative, way of phrasing and feedback. That is why a proposition for an accordingly designed dialogue model will be presented. Thirdly, the increase in demand to modelling must be handled. For this we propose a method of decoupling model and dialogue control. Finally, we will present a proposition for handling dialogue control, that makes adaptivity scalable and generically manageable.