Building Machine Translation Systems for the Next Thousand Languages

TL;DR

This paper reports on developing practical machine translation systems for over a thousand languages, focusing on dataset creation, model training, and evaluation challenges for underrepresented languages.

Contribution

It introduces methods for dataset collection, multilingual model training, and analysis of evaluation metrics for a vast number of languages, advancing inclusive machine translation.

Findings

Successful creation of datasets for 1500+ languages

Development of multilingual models for over 100 high-resource languages and 1000+ low-resource languages

Identification of limitations in current evaluation metrics and common error modes

Abstract

In this paper we share findings from our effort to build practical machine translation (MT) systems capable of translating across over one thousand languages. We describe results in three research domains: (i) Building clean, web-mined datasets for 1500+ languages by leveraging semi-supervised pre-training for language identification and developing data-driven filtering techniques; (ii) Developing practical MT models for under-served languages by leveraging massively multilingual models trained with supervised parallel data for over 100 high-resource languages and monolingual datasets for an additional 1000+ languages; and (iii) Studying the limitations of evaluation metrics for these languages and conducting qualitative analysis of the outputs from our MT models, highlighting several frequent error modes of these types of models. We hope that our work provides useful insights to practitioners working towards building MT systems for currently understudied languages, and highlights research directions that can complement the weaknesses of massively multilingual models in data-sparse settings.