Traditionally, dialogue moves (or speech acts) are defined using preconditions, effects, and a decomposition [Allen1987]. From the perspective of implementing a dialogue move engine, we think it may be useful to think about what a dialogue system (or any dialogue participant) actually needs to do (not necessarily in a sequential order):
Instead of defining the dialogue moves themselves in terms of preconditions and effects, we define update rules (u-rules) and selection rules (s-rules) for updating the DIS based on the recognised move(s) and selecting the next move(s), respectively.
The update rules are rules that update the (dynamic) information state, e.g. when the user has input something to the system. The selection rules are rules that both update the (dynamic) information state and selects a dialogue move to be executed by the system. Both rule types have preconditions and effects. The preconditions are a list of conditions that must be true of the information state. The effects are a list of operations to be executed if the preconditions are true. The preconditions must guarantee that the effects can be executed.
Dialogue move definitions consist of a name, a type (optional) and a list of number and types of arguments (e.g., speaker, content, etc). Dialogue moves are the output of analysis and input to generation. Also, they are the objects selected by s-rules. U-rules may refer to them, and they may be part of the information state.
We also use the term tacit move to refer to the act of applying an update rule, i.e. the act of updating the DIS.