Does it care what you asked? Understanding Importance of Verbs in Deep Learning QA System

TL;DR

This paper investigates the role of verbs in deep learning QA systems, revealing that verbs often have little impact on system decisions due to dataset characteristics and model attention mechanisms.

Contribution

It uncovers the limited influence of verbs in QA decisions and analyzes the neural network's attention and hidden layers to explain this phenomenon, highlighting dataset biases.

Findings

Swapping verbs with antonyms rarely changes system output

Verbs have minimal influence in over 90% of cases

Dataset characteristics contribute to the observed phenomenon

Abstract

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.