Automated Classification of Computing Education Questions using Bloom’s Taxonomy

Abstract

Bloom’s taxonomy is a well-known and widely used method of classifying assessment tasks. However, the application of Bloom’s taxonomy in computing education is often difficult and the classification often suffers from poor inter-rater reliability. Automated approaches using machine learning techniques show potential, but their performance is limited by the quality and quantity of the training set. We implement a machine learning model to classify programming questions according to Bloom’s taxonomy using Google’s BERT as the base model, and the Canterbury QuestionBank as a source of questions categorised by computing education experts. Our results demonstrate that the model was able to successfully predict the categories with moderate success, but was more successful in categorising questions at the lower levels of Bloom’s taxonomy. This work demonstrates the potential for machine learning to assist teachers in the analysis of assessment items.