Automatic Grading of Short Answers for MOOC via Semi-supervised Document Clustering.
Shumin Jing · Educational Data Mining · 2015
Developing an effective and impartial grading system for short answers is a challenging problem in educational measurement and assessment, due to the diversity of answers and the subjectivity of graders. In this paper, we design an automatic grading approach for short answers, based on the non-negative semi-supervised document clustering method. After assigning several answer keys, our approach is able to group the large amount of short answers into multiple sets, and output the score for each answer automatically. In this manner, the effort of teachers can be greatly reduced. Moreover, our approach allows the interaction with teachers, and therefore the system performance could be further enhanced. Experimental results on two datasets demonstrate the effectiveness of our approach.