Leveraging Convolutional Neural Networks for Enhanced Educational Question Classification

• Published in: SN computer science Vol. 5; no. 8; p. 1052
• Main Authors: Deena, G., Ancy Breen, W., Raja, K., Shanmuganathan, C.
• Format: Journal Article
• Language: English
• Published: Singapore Springer Nature Singapore 01.12.2024; Springer Nature B.V
• Subjects: Accuracy; Advances in Computational Approaches for Image Processing; Artificial neural networks; Automatic classification; Automation; Blooms taxonomy; Classification; Cloud Applications and Network Security; Cognitive ability; Computer Imaging; Computer Science; Computer Systems Organization and Communication Networks; Critical thinking; Data Structures and Information Theory; Educational objectives; Feature selection; Information Systems and Communication Service; Keywords; Knowledge; Language; Machine learning; Methods; Natural language processing; Original Research; Pattern Recognition and Graphics; Regression models; Skills; Software Engineering/Programming and Operating Systems; Students; Support vector machines; Taxonomy; Vision; Wireless Networks; Convolutional Neural Network; Word embedding; Bloom’s Taxonomy; Machine Learning
• ISSN: 2661-8907; 2662-995X; 2661-8907
• DOI: 10.1007/s42979-024-03428-6

Question classification is a critical task in natural language processing, as well as applications in question-answering systems and intelligent personal assistants. This study presents novel machine learning approach for automatic question classification based on Bloom’s taxonomy, a hierarchical structure of cognitive levels. Traditional methods rely on feature engineering and classic machine learning models, but these methods often lack the ability to understand complex patterns and connections in natural language data. This study proposes a novel approach using a convolutional neural network (CNN) model, which can automatically learn relevant features from input data without extensive feature engineering. The model classifies questions into pre-established categories or labels using an embedding layer, convolutional and pooling layers, and a fully linked layer for classification. We assess the model’s efficacy on a publicly accessible dataset generated by the subject expert and compare it to conventional machine learning methods like logistic regression and support vector machines. The CNN-based model outperforms conventional techniques, achieving greater recall, accuracy, precision, and F1-score.