Essay Question Generator based on Bloom’s Taxonomy for Assessing Automated Essay Scoring System

Abstract

An automated essay scoring system (AES) is advantageous in evaluating student’s learning outcomes since it gives them the chance to exhibit their knowledge. Most of the AES is using machine learning (ML) to enhance student’s scores but did not consider the proper construction of the essay questions. This study aims to integrate the cognitive level of Blooms’ taxonomy (BT) in constructing essay questions and compare the scores of the student. Identifying the most appropriate ML method in classifying essay exam questions (EEQ) based on BT that will be embedded in the Essay Question Generator (EQG). Using F1-Measure, the evaluation results show that the Support Vector Machine (SVM) (85.7%) outperforms Naïve Bayes (82.6%) and K-Nearest Neighbor (77.6%). Therefore, SVM together with the NLP techniques is applied to automatically extract essay questions from the given text for the teachers to select and apply. The EQG was evaluated using the scores of 375 students who answered two sets of essay exam questions using Bloom’s Taxonomy (BT) and without Bloom’s taxonomy (NBT). Using frequency distribution, the scores between two types were evaluated and the result shows that most students performed well in answering the essay exam using BT 5.6% of the students obtains a perfect score of 5.0 but nobody got 5.0 for NBT. In a conclusion, this study shows that the essay questions constructed according to BT cognitive level produce higher scores using EQG when compared to exam questions prepared by the teachers.