Automatic Question Generation Using Natural Language Processing and Transformers

Abstract

Online education's rapid growth and the rise of E-learning tools have raised the demand for creating assessments and challenging questions for learners which require significant effort to create suitable content for testing. Consequently, automatic question generation aims to automate the creation of different types of questions in the shortest period of time possible by applying minimal human effort all using natural language processing and transformers. This paper proposes different methodologies to generate questions like true or false, fill in the blanks, matching, multiple-choice, and "Wh-" questions specified from a given context. Transformer models including GPT, T5, and BERT are used to achieve the methodologies needed to generate such questions. The system, tested through surveys, generated an 80% satisfaction rate among teacher participants and showed potential to generate questions similar to ChatGPT's.