Towards Example-Based NMT with Multi-Levenshtein Transformers
This work addresses the need for more transparent translation decisions in machine translation, though it is incremental as it builds on existing RAMT and Levenshtein Transformer methods.
The paper tackles the problem of improving transparency in Retrieval-Augmented Machine Translation (RAMT) by proposing a novel architecture that edits multiple fuzzy matches from memory, resulting in increased translation scores and more target spans copied from examples.
Retrieval-Augmented Machine Translation (RAMT) is attracting growing attention. This is because RAMT not only improves translation metrics, but is also assumed to implement some form of domain adaptation. In this contribution, we study another salient trait of RAMT, its ability to make translation decisions more transparent by allowing users to go back to examples that contributed to these decisions. For this, we propose a novel architecture aiming to increase this transparency. This model adapts a retrieval-augmented version of the Levenshtein Transformer and makes it amenable to simultaneously edit multiple fuzzy matches found in memory. We discuss how to perform training and inference in this model, based on multi-way alignment algorithms and imitation learning. Our experiments show that editing several examples positively impacts translation scores, notably increasing the number of target spans that are copied from existing instances.