Boosting the Value of Online Class Forum Facilities

Abstract

We provide experimental results that help automate two related tasks around online instructional forum use. We showhow forum posts can automatically be evaluated for inclusion in a course specific frequently asked questions list. Wefurther show how computer algorithms can help instructorsidentify students who deserve course credit for their participation in the class forum. We deploy forum usage statisticsthat are readily available in most online forum facilities totrain two algorithms. For the FAQ inclusion task we train alogistic regression to an accuracy of 0.78. As ground truthwe asked teaching assistants of one course to provide expertjudgment for use in the regression training phase. For theacademic credit apportioning task we again used experts tocompare pairs of forum contributions. Using those judgmentswe trained a support vector machine to approximate the resulting rankings of posts. We reached a ranking accuracy of 76%, where accuracy reflects minimization of faulty pairwise rankings compared to the expert gold set.