Adversarial Examples for Evaluating Reading Comprehension Systems

Abstract

Standard accuracy metrics indicate that reading comprehension systems are making rapid progress, but the extent to which these systems truly understand language remains unclear. To reward systems with real language understanding abilities, we propose an adversarial evaluation scheme for the Stanford Question Answering Dataset (SQuAD). Our method tests whether systems can answer questions about paragraphs that contain adversarially inserted sentences, which are automatically generated to distract computer systems without changing the correct answer or misleading humans. In this adversarial setting, the accuracy of sixteen published models drops from an average of $75\%$ F1 score to $36\%$; when the adversary is allowed to add ungrammatical sequences of words, average accuracy on four models decreases further to $7\%$. We hope our insights will motivate the development of new models that understand language more precisely.

EMNLP 2017

References in corpus (6)

• Machine Comprehension Using Match-LSTM and Answer Pointer
• Simple Black-Box Adversarial Perturbations for Deep Networks
• Learning Recurrent Span Representations for Extractive Question Answering
• Multi-Perspective Context Matching for Machine Comprehension
• End-to-End Answer Chunk Extraction and Ranking for Reading Comprehension
• Exploring Question Understanding and Adaptation in Neural-Network-Based Question Answering

Cited by in corpus (4)

• Deep Learning: A Critical Appraisal
• DANCin SEQ2SEQ: Fooling Text Classifiers with Adversarial Text Example Generation
• Mapping to Declarative Knowledge for Word Problem Solving
• Analyzing Language Learned by an Active Question Answering Agent