Abstract
Over the past decade, the achievement of Course Learning Outcomes (CLOs) have become the cornerstone for ensuring the quality of graduates in higher education institutions. In practice, the formulation of the appropriate course learning outcomes along with the teaching strategies and assessment methods support the high achievement of students learning. Although, most of the accreditation agencies locally and internationally provide clear guidance about the program learning outcomes, the investigation of the course learning outcomes has received less attention. In addition, the establishment of the appropriate course learning outcomes according to the required level of learning and course requirements is considered as an issue of big challenge and complexity. In practice, the challenge is evident through (1) the big variations of CLOs between similar courses in similar disciplines and (2) the inappropriateness of CLOs with respect to the level of required learning at the course and program levels. In this paper, we propose a novel approach to evaluate the quality and compatibility of CLOs against a set of ideal course learning outcomes that meet well-defined measurement criteria of good CLOs. In doing so, a set of CLOs for core courses in several disciplines have been collected and prepared according the criteria of good CLOs. We apply machine learning methods to rank the learning outcomes against a gold standard set of ideal CLOs. We use a dependency parser to parse the text of the learning outcomes and find similar words through a word embedding model which are fed it to our decision tree built from the gold standard to measure the quality of the new unseen CLOs. The results of our approach show very impressive results in measuring the quality of new CLOs against a set of standard CLOs.