Integrating Translation Memories into Non-Autoregressive Machine Translation

TL;DR

This paper introduces TM-LevT, a novel non-autoregressive translation model that effectively integrates Translation Memories, achieving autoregressive-level performance with reduced decoding complexity.

Contribution

The paper proposes TM-LevT, a new variant of Levenshtein Transformer that incorporates TMs and modifies training for improved translation quality in non-autoregressive models.

Findings

TM-LevT matches autoregressive performance

Incorporating TMs reduces decoding load

Training modifications improve translation quality

Abstract

Non-autoregressive machine translation (NAT) has recently made great progress. However, most works to date have focused on standard translation tasks, even though some edit-based NAT models, such as the Levenshtein Transformer (LevT), seem well suited to translate with a Translation Memory (TM). This is the scenario considered here. We first analyze the vanilla LevT model and explain why it does not do well in this setting. We then propose a new variant, TM-LevT, and show how to effectively train this model. By modifying the data presentation and introducing an extra deletion operation, we obtain performance that are on par with an autoregressive approach, while reducing the decoding load. We also show that incorporating TMs during training dispenses to use knowledge distillation, a well-known trick used to mitigate the multimodality issue.