Emergent Translation in Multi-Agent Communication

TL;DR

This paper introduces a multi-agent communication framework where agents learn to translate languages through visual grounding and interaction, without relying on parallel corpora, leading to emergent translation capabilities.

Contribution

The work demonstrates that language translation can emerge from grounded multimodal interactions in a multi-agent setting, bypassing the need for parallel data.

Findings

Emergent translation arises from visual grounding in a multi-agent environment.

The model outperforms baselines on word and sentence translation tasks.

Multilingual communities improve translation learning speed and quality.

Abstract

While most machine translation systems to date are trained on large parallel corpora, humans learn language in a different way: by being grounded in an environment and interacting with other humans. In this work, we propose a communication game where two agents, native speakers of their own respective languages, jointly learn to solve a visual referential task. We find that the ability to understand and translate a foreign language emerges as a means to achieve shared goals. The emergent translation is interactive and multimodal, and crucially does not require parallel corpora, but only monolingual, independent text and corresponding images. Our proposed translation model achieves this by grounding the source and target languages into a shared visual modality, and outperforms several baselines on both word-level and sentence-level translation tasks. Furthermore, we show that agents in a multilingual community learn to translate better and faster than in a bilingual communication setting.