A Reinforced Generation of Adversarial Examples for Neural Machine Translation

Abstract

Neural machine translation systems tend to fail on less decent inputs despite its significant efficacy, which may significantly harm the credibility of this systems-fathoming how and when neural-based systems fail in such cases is critical for industrial maintenance. Instead of collecting and analyzing bad cases using limited handcrafted error features, here we investigate this issue by generating adversarial examples via a new paradigm based on reinforcement learning. Our paradigm could expose pitfalls for a given performance metric, e.g., BLEU, and could target any given neural machine translation architecture. We conduct experiments of adversarial attacks on two mainstream neural machine translation architectures, RNN-search, and Transformer. The results show that our method efficiently produces stable attacks with meaning-preserving adversarial examples. We also present a qualitative and quantitative analysis for the preference pattern of the attack, demonstrating its capability of pitfall exposure.

arXiv comment12 pages, ACL2020

References in corpus (13)

• Neural Machine Translation by Jointly Learning to Align and Translate
• High-Dimensional Continuous Control Using Generalized Advantage Estimation
• Generating Natural Adversarial Examples
• Synthetic and Natural Noise Both Break Neural Machine Translation
• An Actor-Critic Algorithm for Sequence Prediction
• On Adversarial Examples for Character-Level Neural Machine Translation
• Adafactor: Adaptive Learning Rates with Sublinear Memory Cost
• Seq2Sick: Evaluating the Robustness of Sequence-to-Sequence Models with Adversarial Examples
• Robust Neural Machine Translation with Doubly Adversarial Inputs
• Towards Robust Neural Machine Translation
• A Study of Reinforcement Learning for Neural Machine Translation
• On Evaluation of Adversarial Perturbations for Sequence-to-Sequence Models
• Exploring the Robustness of NMT Systems to Nonsensical Inputs