A Model of Dialog Based on Fuzzy Set Concept

Abstract

Dialog is considered as the exchange of information between two persons who speak natural language to und e rstand each other. However natural language is so ambiguous both in expression an d logi c that for understanding it is very important to supplement the information through so me r e as o ning or imagination. In this paper, we analyze this thinking process using fuzzy se ts theory and approach a modelling of dialog. The dialog in the model is characterized by the set o f objects talked about, the set of their attributes, etc. There are two important fa c tor s in th e p ro ce ss of dialog: fuzzy information on the objects given by one person and the o ther person's fuzzy knowledge on the o bjects. Both fuzzy information and knowledge are expresse d by fuzzy se ts. The main problem in the paper is to compare these two fuzzy sets with each o th e r. For this p urpose the concept of truth qualification in fuzzy logic is useful. The authors propose a generalized truth value instead of an ordinary one which appears in Zadeh's truth qualification rule. Then one of the two persons, say a listener in the dialog, can unders t ano by r e f e rring the generalized truth value what the other wants to communicate.