Using Machine Translation to Augment Multilingual Classification

TL;DR

This paper investigates using machine translation to generate training data for multilingual classifiers and introduces a novel loss technique to mitigate translation-related issues, improving model tuning across languages.

Contribution

It demonstrates the effectiveness of machine translation for multilingual classification and introduces a new loss method to enhance training on translated data.

Findings

Translated data are sufficiently high quality for model tuning.

The novel loss technique improves model performance on translated data.

Machine translation can reduce annotation effort for multilingual tasks.

Abstract

An all-too-present bottleneck for text classification model development is the need to annotate training data and this need is multiplied for multilingual classifiers. Fortunately, contemporary machine translation models are both easily accessible and have dependable translation quality, making it possible to translate labeled training data from one language into another. Here, we explore the effects of using machine translation to fine-tune a multilingual model for a classification task across multiple languages. We also investigate the benefits of using a novel technique, originally proposed in the field of image captioning, to account for potential negative effects of tuning models on translated data. We show that translated data are of sufficient quality to tune multilingual classifiers and that this novel loss technique is able to offer some improvement over models tuned without it.