Towards Generalized Methods for Automatic Question Generation in Educational Domains

Students learn more from doing activities and practicing their skills on assessments, yet it can be challenging and time consuming to generate such practice opportunities. In our work, we examine how advances in natural language processing and question generation may help address this issue. In particular, we present a pipeline for generating and evaluating questions from text-based learning materials in an introductory data science course. The pipeline includes applying a text-to-text transformer (T5) question generation model and a concept hierarchy extraction model on the text content, then scoring the generated questions based on their relevance to the extracted key concepts. We further evaluated the question quality with three different approaches: information score, automated rating by a trained model (Google GPT-3) and manual review by human instructors. Our results showed that the generated questions were rated favorably by all three evaluation methods. We conclude with a discussion of the strengths and weaknesses of the generated questions and outline the next steps towards refining the pipeline and promoting natural language processing research in educational domains.