In the following, we show how the specifications of domain and task model are used by the system for dialogue processing. In order to do so, we do not need to rely on any prestructured dialogue model such as dialogue grammars or finite-state automata. Moreover, we show that although determining the discourse relations may rely on domain-specific knowledge the formulation of the algorithms is domain-independent. Consequently, one particular dialogue strategy can be used in different domains. We understand by dialogue strategy any sequence of actions undertaken by the dialogue system in order to obtain information towards one or more goals. We understand by goal any amount of information sufficient for the system to perform one of the tasks it has been designed to perform. By saying `a task is executed' we refer to the state of the system in which (a) enough information for a goal to be identified uniquely and (b) all required information necessary to perform the actions associated with the goal have been gathered.
As long as a user is engaged with the system in a dialogue, it is then the task of the dialogue system
This section describes how this strategy has been realized.