Measuring College Students' Multidisciplinary Learning: A Novel Application of Natural Language Processing

Using data from approximately 342,000 course-taking records collected from 4406 college students enrolled at Taipei Tech during the 2009-2012 academic years, we examine the impact of multidisciplinarity on students' academic performance. Our study contributes to the literature in three ways. First, by applying natural language processing (NLP), we analyze course descriptions of 375 subject areas from the Classification of Instructional Programs and measure the pairwise distances among them. Second, based on the course-taking records and the subject area distribution, we measure each student's degree of multidisciplinary learning using a proposed weighted entropy formula. Third, using the proposed multidisciplinary index, we find that the impact of multidisciplinary course-taking experience on individual students' academic performance varies across academic fields. In the college of engineering, the college of electrical engineering and computer science, and the college of mechanical and electrical engineering, a higher level of multidisciplinarity is associated with a higher average weighted GPA in core courses. However, a positive association does not exist for students in the college of management.