Learn To Remember: Transformer with Recurrent Memory for Document-Level Machine Translation

TL;DR

This paper introduces a recurrent memory mechanism into the Transformer architecture to enhance document-level machine translation by maintaining context across sentences, leading to improved translation coherence and performance.

Contribution

The paper proposes a novel recurrent memory unit for Transformers, enabling better context integration for document-level translation without increasing computational complexity significantly.

Findings

Achieved an average of 0.91 s-BLEU improvement over sentence-level baseline.

Set new state-of-the-art results on TED and News datasets.

Demonstrated effective context modeling with recurrent memory in Transformer.

Abstract

The Transformer architecture has led to significant gains in machine translation. However, most studies focus on only sentence-level translation without considering the context dependency within documents, leading to the inadequacy of document-level coherence. Some recent research tried to mitigate this issue by introducing an additional context encoder or translating with multiple sentences or even the entire document. Such methods may lose the information on the target side or have an increasing computational complexity as documents get longer. To address such problems, we introduce a recurrent memory unit to the vanilla Transformer, which supports the information exchange between the sentence and previous context. The memory unit is recurrently updated by acquiring information from sentences, and passing the aggregated knowledge back to subsequent sentence states. We follow a two-stage training strategy, in which the model is first trained at the sentence level and then finetuned for document-level translation. We conduct experiments on three popular datasets for document-level machine translation and our model has an average improvement of 0.91 s-BLEU over the sentence-level baseline. We also achieve state-of-the-art results on TED and News, outperforming the previous work by 0.36 s-BLEU and 1.49 d-BLEU on average.