Improving Multilingual Translation by Representation and Gradient Regularization

Yilin Yang, Akiko Eriguchi, Alexandre Muzio, Prasad Tadepalli, Stefan Lee, Hany Hassan


Abstract
Multilingual Neural Machine Translation (NMT) enables one model to serve all translation directions, including ones that are unseen during training, i.e. zero-shot translation. Despite being theoretically attractive, current models often produce low quality translations – commonly failing to even produce outputs in the right target language. In this work, we observe that off-target translation is dominant even in strong multilingual systems, trained on massive multilingual corpora. To address this issue, we propose a joint approach to regularize NMT models at both representation-level and gradient-level. At the representation level, we leverage an auxiliary target language prediction task to regularize decoder outputs to retain information about the target language. At the gradient level, we leverage a small amount of direct data (in thousands of sentence pairs) to regularize model gradients. Our results demonstrate that our approach is highly effective in both reducing off-target translation occurrences and improving zero-shot translation performance by +5.59 and +10.38 BLEU on WMT and OPUS datasets respectively. Moreover, experiments show that our method also works well when the small amount of direct data is not available.
Anthology ID:
2021.emnlp-main.578
Volume:
Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing
Month:
November
Year:
2021
Address:
Online and Punta Cana, Dominican Republic
Editors:
Marie-Francine Moens, Xuanjing Huang, Lucia Specia, Scott Wen-tau Yih
Venue:
EMNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
7266–7279
Language:
URL:
https://aclanthology.org/2021.emnlp-main.578
DOI:
10.18653/v1/2021.emnlp-main.578
Bibkey:
Cite (ACL):
Yilin Yang, Akiko Eriguchi, Alexandre Muzio, Prasad Tadepalli, Stefan Lee, and Hany Hassan. 2021. Improving Multilingual Translation by Representation and Gradient Regularization. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 7266–7279, Online and Punta Cana, Dominican Republic. Association for Computational Linguistics.
Cite (Informal):
Improving Multilingual Translation by Representation and Gradient Regularization (Yang et al., EMNLP 2021)
Copy Citation:
PDF:
https://aclanthology.org/2021.emnlp-main.578.pdf
Video:
 https://aclanthology.org/2021.emnlp-main.578.mp4
Code
 yilinyang7/fairseq_multi_fix
Data
OPUS-100