Automated Mapping of Course Outcomes to Program Outcomes using Natural Language Processing and Machine Learning

Curriculum mapping is an essential academic requirement in outcome-based education (OBE) for successfully implementing the academic program. It is a monitoring tool for course coordinators/instructors to align their curricula according to corresponding program outcomes. This mapping is commonly represented by a two-dimensional matrix depicting the relationship between the Course Outcomes (CO) and Program Outcomes (PO), popularly known as CO-PO mapping. This mapping is a challenging task, even for experienced teachers, as it entails knowledge of the standards and practical aspects of effective mapping, which are not well-documented or readily available. Consequently, this paper aims to tackle this challenge by proposing an AI-based automated system for curriculum mapping, which allows teachers to inspect the CO-PO mappings, identify shortcomings, and recommending suggestions for enhancing curriculum mapping. A dataset is prepared by collecting data from publicly available CO-PO mapped curriculums of the Computer Science Engineering program of seven Indian Universities. Since one CO can be mapped to multiple POs, this work employs machine learning and deep learning models with NLP to solve this multi-label classification task. An empirical study is performed to compare the efficiency of different feature extraction and classification methods. The results indicate that the BERT-based model outperforms the other models in the accuracy of mapping.