Adversarial Neural Machine Translation

Lijun Wu, Yingce Xia, Fei Tian, Li Zhao, Tao Qin, Jianhuang Lai, Tie-Yan Liu
Proceedings of The 10th Asian Conference on Machine Learning, PMLR 95:534-549, 2018.

Abstract

In this paper, we study a new learning paradigm for neural machine translation (NMT). Instead of maximizing the likelihood of the human translation as in previous works, we minimize the distinction between human translation and the translation given by an NMT model. To achieve this goal, inspired by the recent success of generative adversarial networks (GANs), we employ an adversarial training architecture and name it as Adversarial-NMT. In Adversarial-NMT, the training of the NMT model is assisted by an adversary, which is an elaborately designed $2D$ convolutional neural network (CNN). The goal of the adversary is to differentiate the translation result generated by the NMT model from that by human. The goal of the NMT model is to produce high quality translations so as to cheat the adversary. A policy gradient method is leveraged to co-train the NMT model and the adversary. Experimental results on English$\rightarrow$French and German$\rightarrow$English translation tasks show that Adversarial-NMT can achieve significantly better translation quality than several strong baselines.

Cite this Paper

BibTeX
@InProceedings{pmlr-v95-wu18a, title = {Adversarial Neural Machine Translation}, author = {Wu, Lijun and Xia, Yingce and Tian, Fei and Zhao, Li and Qin, Tao and Lai, Jianhuang and Liu, Tie-Yan}, booktitle = {Proceedings of The 10th Asian Conference on Machine Learning}, pages = {534--549}, year = {2018}, editor = {Zhu, Jun and Takeuchi, Ichiro}, volume = {95}, series = {Proceedings of Machine Learning Research}, month = {14--16 Nov}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v95/wu18a/wu18a.pdf}, url = {https://proceedings.mlr.press/v95/wu18a.html}, abstract = {In this paper, we study a new learning paradigm for neural machine translation (NMT). Instead of maximizing the likelihood of the human translation as in previous works, we minimize the distinction between human translation and the translation given by an NMT model. To achieve this goal, inspired by the recent success of generative adversarial networks (GANs), we employ an adversarial training architecture and name it as Adversarial-NMT. In Adversarial-NMT, the training of the NMT model is assisted by an adversary, which is an elaborately designed $2D$ convolutional neural network (CNN). The goal of the adversary is to differentiate the translation result generated by the NMT model from that by human. The goal of the NMT model is to produce high quality translations so as to cheat the adversary. A policy gradient method is leveraged to co-train the NMT model and the adversary. Experimental results on English$\rightarrow$French and German$\rightarrow$English translation tasks show that Adversarial-NMT can achieve significantly better translation quality than several strong baselines.} }
Endnote
%0 Conference Paper %T Adversarial Neural Machine Translation %A Lijun Wu %A Yingce Xia %A Fei Tian %A Li Zhao %A Tao Qin %A Jianhuang Lai %A Tie-Yan Liu %B Proceedings of The 10th Asian Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2018 %E Jun Zhu %E Ichiro Takeuchi %F pmlr-v95-wu18a %I PMLR %P 534--549 %U https://proceedings.mlr.press/v95/wu18a.html %V 95 %X In this paper, we study a new learning paradigm for neural machine translation (NMT). Instead of maximizing the likelihood of the human translation as in previous works, we minimize the distinction between human translation and the translation given by an NMT model. To achieve this goal, inspired by the recent success of generative adversarial networks (GANs), we employ an adversarial training architecture and name it as Adversarial-NMT. In Adversarial-NMT, the training of the NMT model is assisted by an adversary, which is an elaborately designed $2D$ convolutional neural network (CNN). The goal of the adversary is to differentiate the translation result generated by the NMT model from that by human. The goal of the NMT model is to produce high quality translations so as to cheat the adversary. A policy gradient method is leveraged to co-train the NMT model and the adversary. Experimental results on English$\rightarrow$French and German$\rightarrow$English translation tasks show that Adversarial-NMT can achieve significantly better translation quality than several strong baselines.
APA
Wu, L., Xia, Y., Tian, F., Zhao, L., Qin, T., Lai, J. & Liu, T.. (2018). Adversarial Neural Machine Translation. Proceedings of The 10th Asian Conference on Machine Learning, in Proceedings of Machine Learning Research 95:534-549 Available from https://proceedings.mlr.press/v95/wu18a.html.

Related Material