Multilingual Pixel Representations for Translation and Effective Cross-lingual Transfer

TL;DR

This paper presents a novel approach to multilingual machine translation using pixel representations, which improve transferability and data efficiency across diverse languages and scripts compared to traditional subword embeddings.

Contribution

The paper introduces pixel representations for multilingual translation, demonstrating their advantages in cross-lingual transfer and data efficiency over subword-based methods.

Findings

Pixel representations enable effective transfer to unseen scripts.

They are more data-efficient than vocabulary expansion methods.

Parameter sharing within and across scripts enhances performance.

Abstract

We introduce and demonstrate how to effectively train multilingual machine translation models with pixel representations. We experiment with two different data settings with a variety of language and script coverage, demonstrating improved performance compared to subword embeddings. We explore various properties of pixel representations such as parameter sharing within and across scripts to better understand where they lead to positive transfer. We observe that these properties not only enable seamless cross-lingual transfer to unseen scripts, but make pixel representations more data-efficient than alternatives such as vocabulary expansion. We hope this work contributes to more extensible multilingual models for all languages and scripts.