# Domain Adaptive Inference for Neural Machine Translation

## TL;DR

This paper introduces a novel domain adaptive inference method for neural machine translation that improves performance on new domains without losing accuracy on original domains, using ensemble weighting and Bayesian interpolation techniques.

## Contribution

It proposes a new adaptive ensemble decoding scheme based on Bayesian Interpolation extended with source info, enhancing domain adaptation in NMT.

## Key findings

- Strong improvements across multiple test domains
- Effective adaptation without domain labels
- Comparison of different fine-tuning methods

## Abstract

We investigate adaptive ensemble weighting for Neural Machine Translation, addressing the case of improving performance on a new and potentially unknown domain without sacrificing performance on the original domain. We adapt sequentially across two Spanish-English and three English-German tasks, comparing unregularized fine-tuning, L2 and Elastic Weight Consolidation. We then report a novel scheme for adaptive NMT ensemble decoding by extending Bayesian Interpolation with source information, and show strong improvements across test domains without access to the domain label.