A Class of Fuzzy Featural Models of Similarity Judgments

Abstract

Tverksy’s [10] seminal paper on feature-matching models for similarity judgments has spawned a number of generalized models (e.g., [2] and [3]) which purport to handle asymmetric similarity data (i.e., where the statement “A is like B” elicits a different degree of judged similarity or truth-value than “B is like A”). The same period has seen rapid development in so-called additive models of clustering based on the representation of similarities as combinations of overlapping discrete features as in [7]. The fuzzy clustering literature, on the other hand, has not dealt explicitly with problems of asymmetric data and it has developed almost independently of psychological similarity judgment models (cf. for instance, [1], [4], and [6]). Moreover, their algorithms are distance-based and spatially oriented rather than tree or feature oriented.