English

Context-Aware Cross-Attention for Non-Autoregressive Translation

Computation and Language 2020-11-03 v1

Abstract

Non-autoregressive translation (NAT) significantly accelerates the inference process by predicting the entire target sequence. However, due to the lack of target dependency modelling in the decoder, the conditional generation process heavily depends on the cross-attention. In this paper, we reveal a localness perception problem in NAT cross-attention, for which it is difficult to adequately capture source context. To alleviate this problem, we propose to enhance signals of neighbour source tokens into conventional cross-attention. Experimental results on several representative datasets show that our approach can consistently improve translation quality over strong NAT baselines. Extensive analyses demonstrate that the enhanced cross-attention achieves better exploitation of source contexts by leveraging both local and global information.

Keywords

Cite

@article{arxiv.2011.00770,
  title  = {Context-Aware Cross-Attention for Non-Autoregressive Translation},
  author = {Liang Ding and Longyue Wang and Di Wu and Dacheng Tao and Zhaopeng Tu},
  journal= {arXiv preprint arXiv:2011.00770},
  year   = {2020}
}

Comments

To appear in COLING 2020

R2 v1 2026-06-23T19:50:10.457Z