Impact of Domain-Adapted Multilingual Neural Machine Translation in the Medical Domain
This addresses translation accuracy for medical texts in low-resource language pairs like English-Romanian, but it is incremental as it applies existing adaptation techniques to a specific domain.
The study tackled the problem of improving translation quality for low-resource languages in the medical domain by adapting a multilingual neural machine translation model, resulting in the in-domain adapted model outperforming the out-of-domain model in all automatic metrics and reducing terminology errors.
Multilingual Neural Machine Translation (MNMT) models leverage many language pairs during training to improve translation quality for low-resource languages by transferring knowledge from high-resource languages. We study the quality of a domain-adapted MNMT model in the medical domain for English-Romanian with automatic metrics and a human error typology annotation which includes terminology-specific error categories. We compare the out-of-domain MNMT with the in-domain adapted MNMT. The in-domain MNMT model outperforms the out-of-domain MNMT in all measured automatic metrics and produces fewer terminology errors.