A Reinforced Generation of Adversarial Samples for Neural Machine Translation

Neural machine translation systems tend to fail on less de-cent inputs despite its great efficacy, which may greatly harm the credibility of these 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 samples 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 samples. We also present a qualitative and quantitative analysis for the preference pattern of the attack, showing its capability of pitfall exposure.