Automating Educational Assessment with AI: Leveraging Bloom's Taxonomy and Transformer Models for Question Classification

This study aims to leverage artificial intelligence, particularly large language models (LLMs) and several BERT architecture versions, to categorize educational assessment questions according to Bloom's taxonomy. That lays out a coherent framework for classifying cognitive abilities, spanning from fundamental recollection to higher-order processes like judgment and invention. The study examines the efficacy of several BERT-based models in the domain of text categorization. The utilized dataset comprises computer science evaluation questions that have been methodically pre-processed and organized to facilitate precise categorization. The research evaluates the classification efficacy of several BERT variations, including BERTbase, DistilBERT, ALBERT, and RoBERTa, in aligning queries with the corresponding tiers of Bloom's taxonomy. Findings demonstrate that RoBERTa and DistilBERT surpass the conventional BERT-base model for classification accuracy. provides valuable insights for the enhancement of adaptive learning systems and the improvement of educational evaluation methodologies. The report delineates the limits of existing research and suggests future avenues, including the utilization of ensemble learning techniques and the incorporation of more comprehensive contextual data.