A technique on adjusting the system's knowledge of linguistic hedges to human naive thinking

The concept of a linguistic hedge has been successfully formalized in the field of fuzzy set theory. The meanings of linguistic expressions, such as "very tall", have been well expressed by using membership functions. Although various methods for measuring the membership functions have been proposed, the change of the meaning in the real world has not been treated well. In this paper, we introduce the concept of a population type for the perceived data set, and we conduct psychological experiments to determine the membership functions for the linguistic expressions, while changing the stimulus population type. Then, by using the population type concept, we propose a technique for adjusting the system's knowledge of linguistic hedges to new perceived data, and we discuss the usefulness of the method.