Automatic Question Generation: A Comparative Analysis of Rule-Based and Neural Network-Based Models

Questioning is essential for learning and higher-level cognitive abilities. Manual question construction is time-consuming and requires experts. To streamline the process, researchers have proposed various automatic question generation (AQG) approaches for real-time applications in natural language processing (NLP). This paper provides an overview of the latest developments in AQG and discusses the state-of-the-art rule-based and neural network-based models employed in these systems. In addition, this paper incorporates an empirical assessment of the effectiveness of two prominent models, namely the Heilman and Smith (H &S) model and the transformer-based T5 model, using educational resources’ book chapters as the basis for performance evaluation.