Pedagogical Classification Model Based on Machine Learning

Abstract

AbstractThe quality assurance of E-Learning systems can be guaranteed by a quality instructional design and by the definition of learning objectives associated with modules and programs. Learning objectives are central to teaching and learning in many higher education institutions. However, teachers have limited tools to help them reflect on the learning objectives of the course they create. Bloom's taxonomy's cognitive levels are widely used as a reference standard for classifying E-Learning Contents. However, many action verbs in Bloom's taxonomy overlap at various levels of the hierarchy, leading to confusion about the cognitive level expected. Some research papers have addressed the cognitive classification of E-Learning content such as assessment questions or forum texts, but none has addressed learning objectives. This study proposes a method for classifying learning objectives automatically, by extracting features based on a modified TF-IDF-POS to assign a suitable weight for essential words in the learning objective based on Part-Of-Speech (POS). Then, we use different classifiers combined with those features. To address the problem of the absence of annotated learning objectives dataset, we create a dataset of 2400 items. The classification results achieved the highest accuracies for the models combined with TF-IDF-POS. According to the findings of this study, the proposed method is effective in classifying learning objectives using Bloom's taxonomy. KeywordsPedagogical classificationMachine learningTF-IDFPart-Of-SpeechFeature extractionBloom’s taxonomy