A fuzzy-set-theoretic feature model and its application to asymmetric similarity data analysis

Feature representation models are too restricted in that they consider features as dichotomous variables. In an attempt to construct a more general feature model, it is argued that fuzzy set theory (Zadeh, 1965) gives a natural and promising solution to the problem. Using fuzzy set theory in place of ordinary set theory, a fuzzy feature matching model, which is a generalization of Tversky's contrast model (1977) of similarity, is proposed and is applied to the analysis of an asymmetric similarity matrix, which was obtained by asking 42 undergraduates to judge pairwise similarity among eight countries.