Revisiting Simple Domain Adaptation Methods in Unsupervised Neural Machine Translation
This work addresses domain inconsistency issues in UNMT for researchers and practitioners, but it is incremental as it builds on and modifies existing methods.
The paper tackled domain adaptation in unsupervised neural machine translation (UNMT) by empirically analyzing scenarios and revisiting existing methods like batch weighting and fine-tuning, proposing modifications that improved performance on domain-specific language pairs.
Domain adaptation has been well-studied in supervised neural machine translation (SNMT). However, it has not been well-studied for unsupervised neural machine translation (UNMT), although UNMT has recently achieved remarkable results in several domain-specific language pairs. Besides the inconsistent domains between training data and test data for SNMT, there sometimes exists an inconsistent domain between two monolingual training data for UNMT. In this work, we empirically show different scenarios for unsupervised neural machine translation. Based on these scenarios, we revisit the effect of the existing domain adaptation methods including batch weighting and fine tuning methods in UNMT. Finally, we propose modified methods to improve the performances of domain-specific UNMT systems.