Ordered models for concept analysis

Michael Freund (1)

1 Chercheur indépendant

In this paper, I propose a simple ordered model for categoriza-tion theory. When a concept is grasped through an auxiliary set of features, stemming from a description of the concept or from a resemblance to some typical exemplars, categorial membership, typicality and resemblance can be accounted for by means of weak order relations. These orders render precise the notion of an object 'falling more' than another under this concept, being 'a more typical exemplar' or 'resembling more' this concept. The membership and typicality orders can be naturally extended to compound concepts, without the drawbacks that are classically encountered in conjunction theory. Moreover , these orders reveal themselves to be a particularly adequate tool for the resolution of problems linked with category-based induction.

Mots clés: categorization; prototype theory; categorial membership; typi- cality; deep learning; category-based induction; resemblance theory

Domaines: Psychologie

Michael Freund. Ordered models for concept analysis. 2017. ⟨hal-01668227⟩