The Grammar of Emergent Languages

TL;DR

This paper investigates the syntactic properties of emergent languages in referential games using unsupervised grammar induction, revealing conditions under which syntactic structure appears and advocating for UGI as a standard analysis tool.

Contribution

It demonstrates the suitability of UGI techniques for analyzing emergent languages and provides a comprehensive library to facilitate future research in this area.

Findings

Syntactic structure emerges at certain message lengths and vocabulary sizes.

More complex game scenarios are needed for human-like syntax.

UGI techniques are effective for analyzing emergent language structures.

Abstract

In this paper, we consider the syntactic properties of languages emerged in referential games, using unsupervised grammar induction (UGI) techniques originally designed to analyse natural language. We show that the considered UGI techniques are appropriate to analyse emergent languages and we then study if the languages that emerge in a typical referential game setup exhibit syntactic structure, and to what extent this depends on the maximum message length and number of symbols that the agents are allowed to use. Our experiments demonstrate that a certain message length and vocabulary size are required for structure to emerge, but they also illustrate that more sophisticated game scenarios are required to obtain syntactic properties more akin to those observed in human language. We argue that UGI techniques should be part of the standard toolkit for analysing emergent languages and release a comprehensive library to facilitate such analysis for future researchers.