Semi-automatic short answers clustering and grading with K-Means and Keyword Matching algorithms

Appropriate and timely feedback enhances student learning. Such activities are challenging for professors to do in large classes. This paper provides two semi-automated methods for analyzing short responses: keyword matching and unsupervised document grouping. Three types of questions are tested. For fixed answers, keyword matching is most effective. Document clustering, on the other hand, is effective with a query that is more open-ended. This helps teachers to more efficiently grade students. They might be used in MOOCs to provide students with immediate feedback.