General architecture

The general architecture we are assuming is shown in (1).

 

The components in the architecture are the following:

• Information State (IS), divided into Dynamic IS (DIS) and static IS (SIS)
• Interpreter: Takes input utterances from the user and gives interpretations in terms of moves (including semantic content). The interpretation is stored in the DIS.
• Update module: Applies update rules (specified in SIS) to the DIS according to the update algorithm (also specified in SIS)
• Selection module: Selects move(s) using the selection rules and the move selection algorithm specified in SIS. The resulting moves are stored in the DIS. The update module and the selection module together make up the dialogue move engine.
• Generator: Generates ouput utterances based on the contents of the DIS.
• Control: wires together the other modules, either in sequence or through some asynchronous mechanism.

Apart from the general architecture show in (1), the framework also specifies formats for defining update rules, selection rules and dialogue moves (see section 2.2), and provides a set of tools for experimenting with different information states, rules, and algorithms. Simple intepreters and generators are also provided. To build a dialogue system, one needs to supplement the framework with definitions of rules, moves and algorithms, as well as the structure of the dynamic information state.