Semantically characterizing collaborative behavior in an Abstract dialogue framework

Abstract

A fundamental requirement of collaborative dialogue formal systems is ensuring both that all the relevant information will be exposed and also irrelevancies will be avoided. The challenge is to fulfill this requirement in the context of a distributed MAS where each agent is unaware of the private knowledge of the others. We argue that it is possible to give a general treatment to this problem in terms of relevance notions, and propose a partial solution which reduces the problem to that of finding adequate potential relevance notions. Specifically, we present in this work an Abstract Dialogue Framework which provides an environment for studying the behavior of collaborative dialogue systems in terms of abstract relevance notions, together with three Collaborative Semantics each of which defines a different collaborative behavior of the dialogues under the framework. One of these semantics describes an utopian, non practical, behavior which is approximated in different ways by the other two constructive semantics. Complete examples are provided in Propositional Logic Programming. © 2010 Springer Berlin Heidelberg.