Multilingual Multimodal Learning with Machine Translated Text

TL;DR

This paper explores using machine translated English multimodal data to improve multilingual multimodal learning, demonstrating its effectiveness across multiple tasks and languages in the IGLUE benchmark.

Contribution

It introduces the TD-MML framework that leverages machine translation for multilingual multimodal training and proposes metrics to filter low-quality translations.

Findings

Translated data enhances multilingual multimodal learning performance.

Filtering low-quality translations improves model training.

Effective across diverse languages and tasks.

Abstract

Most vision-and-language pretraining research focuses on English tasks. However, the creation of multilingual multimodal evaluation datasets (e.g. Multi30K, xGQA, XVNLI, and MaRVL) poses a new challenge in finding high-quality training data that is both multilingual and multimodal. In this paper, we investigate whether machine translating English multimodal data can be an effective proxy for the lack of readily available multilingual data. We call this framework TD-MML: Translated Data for Multilingual Multimodal Learning, and it can be applied to any multimodal dataset and model. We apply it to both pretraining and fine-tuning data with a state-of-the-art model. In order to prevent models from learning from low-quality translated text, we propose two metrics for automatically removing such translations from the resulting datasets. In experiments on five tasks across 20 languages in the IGLUE benchmark, we show that translated data can provide a useful signal for multilingual multimodal learning, both at pretraining and fine-tuning.