Domain Differential Adaptation for Neural Machine Translation

TL;DR

This paper introduces Domain Differential Adaptation (DDA), a novel strategy for neural machine translation that models domain differences explicitly, leading to improved adaptation performance across various datasets.

Contribution

The paper proposes DDA, a new framework that models domain differences directly rather than smoothing, enhancing neural machine translation adaptation.

Findings

DDA outperforms existing adaptation methods in multiple settings.

Explicit modeling of domain differences improves translation quality.

Experimental results show consistent performance gains.

Abstract

Neural networks are known to be data hungry and domain sensitive, but it is nearly impossible to obtain large quantities of labeled data for every domain we are interested in. This necessitates the use of domain adaptation strategies. One common strategy encourages generalization by aligning the global distribution statistics between source and target domains, but one drawback is that the statistics of different domains or tasks are inherently divergent, and smoothing over these differences can lead to sub-optimal performance. In this paper, we propose the framework of {\it Domain Differential Adaptation (DDA)}, where instead of smoothing over these differences we embrace them, directly modeling the difference between domains using models in a related task. We then use these learned domain differentials to adapt models for the target task accordingly. Experimental results on domain adaptation for neural machine translation demonstrate the effectiveness of this strategy, achieving consistent improvements over other alternative adaptation strategies in multiple experimental settings.