Does it care what you asked? Understanding Importance of Verbs in Deep Learning QA System
This work addresses the problem of robustness in NLP QA systems for researchers, but it is incremental as it builds on existing adversarial example studies.
The study investigated the importance of verbs in a deep learning QA system trained on SQuAD, finding that swapping verbs for antonyms did not change system decisions in over 90% of cases, with the issue traced to dataset characteristics.
In this paper we present the results of an investigation of the importance of verbs in a deep learning QA system trained on SQuAD dataset. We show that main verbs in questions carry little influence on the decisions made by the system - in over 90% of researched cases swapping verbs for their antonyms did not change system decision. We track this phenomenon down to the insides of the net, analyzing the mechanism of self-attention and values contained in hidden layers of RNN. Finally, we recognize the characteristics of the SQuAD dataset as the source of the problem. Our work refers to the recently popular topic of adversarial examples in NLP, combined with investigating deep net structure.