Learning Contextualized Sentence Representations for Document-Level Neural Machine Translation

TL;DR

This paper introduces a framework for enhancing neural machine translation by training models to predict surrounding sentences, leading to more context-aware translations at the document level.

Contribution

It proposes novel methods to learn and incorporate contextualized sentence embeddings into NMT, improving translation quality over existing baselines.

Findings

Both proposed methods significantly improve translation quality.

Pre-training on large monolingual corpora enhances context modeling.

Joint training and pre-training & fine-tuning outperform baseline models.

Abstract

Document-level machine translation incorporates inter-sentential dependencies into the translation of a source sentence. In this paper, we propose a new framework to model cross-sentence dependencies by training neural machine translation (NMT) to predict both the target translation and surrounding sentences of a source sentence. By enforcing the NMT model to predict source context, we want the model to learn "contextualized" source sentence representations that capture document-level dependencies on the source side. We further propose two different methods to learn and integrate such contextualized sentence embeddings into NMT: a joint training method that jointly trains an NMT model with the source context prediction model and a pre-training & fine-tuning method that pretrains the source context prediction model on a large-scale monolingual document corpus and then fine-tunes it with the NMT model. Experiments on Chinese-English and English-German translation show that both methods can substantially improve the translation quality over a strong document-level Transformer baseline.