Finding the Right Recipe for Low Resource Domain Adaptation in Neural Machine Translation

TL;DR

This paper empirically evaluates domain adaptation methods for neural machine translation in low-resource scenarios across multiple domains and languages, providing practical recommendations and datasets.

Contribution

It systematically compares data-centric adaptation techniques in very low and moderately low resource settings, proposing an ensemble approach and releasing relevant datasets and code.

Findings

Ensemble methods improve translation quality in low-resource domains.

Data-centric approaches vary in effectiveness across domains and resource levels.

Practical recommendations for domain adaptation in NMT are provided.

Abstract

General translation models often still struggle to generate accurate translations in specialized domains. To guide machine translation practitioners and characterize the effectiveness of domain adaptation methods under different data availability scenarios, we conduct an in-depth empirical exploration of monolingual and parallel data approaches to domain adaptation of pre-trained, third-party, NMT models in settings where architecture change is impractical. We compare data centric adaptation methods in isolation and combination. We study method effectiveness in very low resource (8k parallel examples) and moderately low resource (46k parallel examples) conditions and propose an ensemble approach to alleviate reductions in original domain translation quality. Our work includes three domains: consumer electronic, clinical, and biomedical and spans four language pairs - Zh-En, Ja-En, Es-En, and Ru-En. We also make concrete recommendations for achieving high in-domain performance and release our consumer electronic and medical domain datasets for all languages and make our code publicly available.