The model of examination question classification based on bloom’s taxonomy using semantic similarity technique

Bloom’s Taxonomy (BT) has generally been used as a guideline in designing a holistic set of examination questions that comprise various cognitive levels. It has been emphasised by Engineering Accreditation Council Malaysia (EAC) and Malaysian Qualifications Agency (MQA) to regulate the quality and standard of education provided by setting the assessment questions aligned with the Course Learning Outcomes (CLO). However, there are inconsistencies in the classification of final examination questions based on Bloom’s Taxonomy. This is because it is manually conducted by academics and is susceptible to discrepancies in the understanding of BT among academics. Most of the research work focused on singlesentence questions that were not based on real examination questions. While previous research has explored examination question classification using a semantic approach, it encountered challenges in achieving high accuracy, which is greater than 80%. Therefore, this research aims to introduce a model to perform examination question classification based on BT using a semantic approach with real examination question and striving to attain an accuracy exceeding 80%. A Question Classification Model (QCM) was developed in this research using Natural Language Processing (NLP) approaches, such as the Stanford POS (Part-Of-Speech) tagger, to preprocess the examination questions into word tokens. Subsequently, Stanford Parser Universal Dependencies (UD) was used to identify the important verbs in the examination questions that reflect the thinking action. This was followed by a comparison between the identified verbs and the list of BT verbs using the WordNet Similarity approach. Moreover, this research has studied, evaluated and enhanced each approach to achieve the best performance for the QCM. Overall, the developed QCM achieved a recorded accuracy rate of 83% in the classification of a set of 200 examination questions based on BT. This research helps to control the assessment quality to meet the classification and fulfil the requirements of Outcome-Based Education (OBE) standards.