A Socially Supported Knowledge-Belief System and Its Application to a Dialogue Parser

摘要

This paper proposes a dynamically-managed knowledge-belief system which can represent continuously-changing beliefs about the world. Each belief in the system has a component reflecting the strength of support from other agents. The system is capable of adapting to a contextual situation due to the continuous revision of belief strengths through interaction with others. As a paradigmatic application of the proposed socially supported knowledge-belief system, a dialogue parser was designed and implemented in Common Lisp CLOS. The parser equipped with a belief management system outputs (a) speaker intention, (b) conveyed meaning, and (c) hearer emotion. Implications of our treatment of intention, meaning, and affective content within an integrated framework are discussed.