Know What You Don't Know: Modeling a Pragmatic Speaker that Refers to Objects of Unknown Categories

TL;DR

This paper introduces a neural pragmatic speaker model for zero-shot reference games that effectively refers to objects of unknown categories, improving communication success by reasoning about uncertainty.

Contribution

It combines zero-shot learning with pragmatic language modeling, extending rational speech act models to handle unknown object categories in visual reference tasks.

Findings

Pragmatic reasoning reduces noun usage compared to literal models.

The model improves reference resolution accuracy.

Fewer distractor categories are named by the pragmatic speaker.

Abstract

Zero-shot learning in Language & Vision is the task of correctly labelling (or naming) objects of novel categories. Another strand of work in L&V aims at pragmatically informative rather than ``correct'' object descriptions, e.g. in reference games. We combine these lines of research and model zero-shot reference games, where a speaker needs to successfully refer to a novel object in an image. Inspired by models of "rational speech acts", we extend a neural generator to become a pragmatic speaker reasoning about uncertain object categories. As a result of this reasoning, the generator produces fewer nouns and names of distractor categories as compared to a literal speaker. We show that this conversational strategy for dealing with novel objects often improves communicative success, in terms of resolution accuracy of an automatic listener.