BloomBERT: A Deep Learning-Based Cognitive Complexity Classifier of Assessment Questions
Jyoti Prakash Meher, Rajib Mall · 2023
We propose a novel automated approach using deep learning techniques to classify assessment questions based on their cognitive complexity. To this end, we first heuristically annotated a dataset of ∼111.6K questions based on Revised Bloom's Taxonomy. Subsequently, we used transfer learning to train an attention-based model for question classification. Our experimental results indicate that our proposed approach outperforms existing classification techniques in terms of F1-Score. Moreover, our proposed model could correctly classify 74.04% of real-life questions.