Automated Answer-Agnostic Diverse Question Generation with Self-Attention Architectures

Abstract

"Automatic generation of questions is a challenging task in Natural Language Processing that has attracted a lot of attention in recent years. Creating a variety of question categories, such as open-ended, true/false, and multiple-choice questions, that cover different aspects of the paragraph’s content remains a challenge. In addition, existing methods have a tendency to generate questions that are identical in structure and content, resulting in a lack of diversity that could impact the accuracy of comprehension assessment. In order to improve comprehension assessment, an innovative approach is required that can generate a wide variety of question types while promoting question diversity."