Exploring Artificial Jabbering for Automatic Text Comprehension Question Generation

Tim Steuer, Anna Filighera, Christoph Rensing

Abstract: Many educational texts lack comprehension questions and authoring them consumes time and money. Thus, in this article, we ask ourselves to what extent artificial jabbering text generation systems can be used to generate textbook comprehension questions. Novel machine learning-based text generation systems jabber on a wide variety of topics with deceptively good performance. To expose the generated texts as such, one often has to understand the actual topic the systems jabbers about. Hence, confronting learners with generated texts may cause them to question their level of knowledge. We built a novel prototype that generates comprehension questions given arbitrary textbook passages. We discuss the strengths and weaknesses of the prototype quantitatively and qualitatively. While our prototype is not perfect, we provide evidence that such systems have great potential as question generators and identify the most promising starting points may leading to (semi) automated generators that support textbook authors and self-studying.