Exploring the Robustness of NMT Systems to Nonsensical Inputs

Abstract

Neural machine translation (NMT) systems have been shown to give undesirable translation when a small change is made in the source sentence. In this paper, we study the behaviour of NMT systems when multiple changes are made to the source sentence. In particular, we ask the following question "Is it possible for an NMT system to predict same translation even when multiple words in the source sentence have been replaced?". To this end, we propose a soft-attention based technique to make the aforementioned word replacements. The experiments are conducted on two language pairs: English-German (en-de) and English-French (en-fr) and two state-of-the-art NMT systems: BLSTM-based encoder-decoder with attention and Transformer. The proposed soft-attention based technique achieves high success rate and outperforms existing methods like HotFlip by a significant margin for all the conducted experiments. The results demonstrate that state-of-the-art NMT systems are unable to capture the semantics of the source language. The proposed soft-attention based technique is an invariance-based adversarial attack on NMT systems. To better evaluate such attacks, we propose an alternate metric and argue its benefits in comparison with success rate.

References in corpus (6)

• Explaining and Harnessing Adversarial Examples
• Adversarial Examples for Evaluating Reading Comprehension Systems
• Pathologies of Neural Models Make Interpretations Difficult
• On Adversarial Examples for Character-Level Neural Machine Translation
• Adversarial Examples for Semantic Segmentation and Object Detection
• Detecting egregious responses in neural sequence-to-sequence models

Cited by in corpus (4)

• Imitation Attacks and Defenses for Black-box Machine Translation Systems
• A Reinforced Generation of Adversarial Examples for Neural Machine Translation
• Smoothed Inference for Adversarially-Trained Models
• Secoco: Self-Correcting Encoding for Neural Machine Translation