English

BERT-JAM: Boosting BERT-Enhanced Neural Machine Translation with Joint Attention

Computation and Language 2020-11-10 v1

Abstract

BERT-enhanced neural machine translation (NMT) aims at leveraging BERT-encoded representations for translation tasks. A recently proposed approach uses attention mechanisms to fuse Transformer's encoder and decoder layers with BERT's last-layer representation and shows enhanced performance. However, their method doesn't allow for the flexible distribution of attention between the BERT representation and the encoder/decoder representation. In this work, we propose a novel BERT-enhanced NMT model called BERT-JAM which improves upon existing models from two aspects: 1) BERT-JAM uses joint-attention modules to allow the encoder/decoder layers to dynamically allocate attention between different representations, and 2) BERT-JAM allows the encoder/decoder layers to make use of BERT's intermediate representations by composing them using a gated linear unit (GLU). We train BERT-JAM with a novel three-phase optimization strategy that progressively unfreezes different components of BERT-JAM. Our experiments show that BERT-JAM achieves SOTA BLEU scores on multiple translation tasks.

Keywords

Cite

@article{arxiv.2011.04266,
  title  = {BERT-JAM: Boosting BERT-Enhanced Neural Machine Translation with Joint Attention},
  author = {Zhebin Zhang and Sai Wu and Dawei Jiang and Gang Chen},
  journal= {arXiv preprint arXiv:2011.04266},
  year   = {2020}
}
R2 v1 2026-06-23T20:00:20.131Z