MDAPT: Multilingual Domain Adaptive Pretraining in a Single Model

TL;DR

This paper investigates how to effectively adapt a single multilingual language model to specific domains across multiple languages, demonstrating that such models can outperform general multilingual models and approach monolingual performance.

Contribution

It introduces techniques for domain adaptive pretraining in a multilingual setting, enabling a single model to excel in domain-specific tasks across multiple languages.

Findings

Multilingual domain-adaptive models outperform general multilingual models.

Single models perform close to monolingual models in domain-specific tasks.

Techniques work across adapter-based and full model pretraining methods.

Abstract

Domain adaptive pretraining, i.e. the continued unsupervised pretraining of a language model on domain-specific text, improves the modelling of text for downstream tasks within the domain. Numerous real-world applications are based on domain-specific text, e.g. working with financial or biomedical documents, and these applications often need to support multiple languages. However, large-scale domain-specific multilingual pretraining data for such scenarios can be difficult to obtain, due to regulations, legislation, or simply a lack of language- and domain-specific text. One solution is to train a single multilingual model, taking advantage of the data available in as many languages as possible. In this work, we explore the benefits of domain adaptive pretraining with a focus on adapting to multiple languages within a specific domain. We propose different techniques to compose pretraining corpora that enable a language model to both become domain-specific and multilingual. Evaluation on nine domain-specific datasets-for biomedical named entity recognition and financial sentence classification-covering seven different languages show that a single multilingual domain-specific model can outperform the general multilingual model, and performs close to its monolingual counterpart. This finding holds across two different pretraining methods, adapter-based pretraining and full model pretraining.