A software platform for evaluating student essays in interdisciplinary learning with topic classification techniques

Abstract

Interdisciplinary learning aims to address the growing need to solve complex problems that go beyond the boundaries of a single discipline. Higher education institutions increasingly recognize the needs and their responsibilities in cultivating university students’ interdisciplinary learning skills. To better facilitate interdisciplinary learning, it is crucial to perform robust evaluations of students’ works. In this project, I have developed an interdisciplinary learning evaluation tool for student essays as an initial step to address this research gap. Equipped with natural language processing techniques and multi-label classification, I trained the tool with a combined public text corpus. The model is evaluated by reading student essays from a digital literacy course in Nanyang Technological University and detecting the number of disciplines presented and the estimated degree of disciplinary integration. The tool demonstrated high classification accuracy and F1- score across all four disciplines and delivers evaluation results similar to human grader. A software application was developed using Python Flask and React JavaScript, and integrated with the machine learning model, to provide student and teacher users with an interface to use the tool and visualize the disciplines analysis.