Estimating the Difficulty of Courses From Syllabuses

Abstract

Syllabuses of many universities are published on the Web, and in some cases, past average scores are also published for the convenience of student registration. In this study, we investigate the relationship between the content of the syllabuses and the average scores, and clarify whether the difficulty level of the courses can be estimated from the syllabuses. In the proposed method, first, the topics that make up the syllabuses are extracted using topic analysis. Then, we find out how much each syllabus contains these topics. The difficulty of courses is estimated from these features using support vector machine. 10-fold cross-validation was performed for the evaluation, and it became clear that the difficulty level can be estimated with an accuracy of 62.3%.