'The red one!': On learning to refer to things based on discriminative properties

摘要

As a first step towards agents learning to communicate about their visual environment, we propose a system that, given visual representations of a referent (CAT) and a context (SOFA), identifies their discriminative attributes, i.e., properties that distinguish them (has_tail). Moreover, although supervision is only provided in terms of discriminativeness of attributes for pairs, the model learns to assign plausible attributes to specific objects (SOFA-has_cushion). Finally, we present a preliminary experiment confirming the referential success of the predicted discriminative attributes.