Adversarial Neural Machine Translation

TL;DR

This paper introduces Adversarial-NMT, a novel training paradigm for neural machine translation that employs adversarial training with a CNN-based discriminator to improve translation quality over traditional likelihood maximization methods.

Contribution

It proposes a new adversarial training framework for NMT inspired by GANs, using a CNN discriminator to enhance translation performance.

Findings

Achieves significantly better translation quality than strong baselines.

Demonstrates effectiveness on English-French and German-English translation tasks.

Introduces a novel adversarial training approach for NMT.

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 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.