Neural Machine Translation (NMT) systems built on multilingual sequence-to-sequence Language Models (msLMs) fail to deliver expected results when the amount of parallel data for a language, as well as the language's representation in the model are limited. This restricts the capabilities of domain-specific NMT systems for low-resource languages (LRLs). As a solution, parallel data from auxiliary domains can be used either to fine-tune or to further pre-train the msLM. We present an evaluation of the effectiveness of these two techniques in the context of domain-specific LRL-NMT. We also explore the impact of domain divergence on NMT model performance. We recommend several strategies for utilizing auxiliary parallel data in building domain-specific NMT models for LRLs.
@article{arxiv.2412.19522,
title = {Exploiting Domain-Specific Parallel Data on Multilingual Language Models for Low-resource Language Translation},
author = {Surangika Ranathungaa and Shravan Nayak and Shih-Ting Cindy Huang and Yanke Mao and Tong Su and Yun-Hsiang Ray Chan and Songchen Yuan and Anthony Rinaldi and Annie En-Shiun Lee},
journal= {arXiv preprint arXiv:2412.19522},
year = {2026}
}