FAQtor : Automatic FAQ generation using online forums

Abstract

Forum content is an accumulation of significant human effort. It is thus sensible to extract prolonged benefit from this investment. We propose a system to automatically generate course specific frequently asked question lists (FAQs) from multi-year forum datasets. The system creates two separate FAQs: a Student FAQ, and an Instructor FAQ. The Student FAQ provides help for queries asked in the past. The Instructor FAQ provides an overview of forum activity in previous course offerings. We present the experimental results for the automated Student FAQ generation from a logistic regression. The classifier reaches an accuracy of 0.78. The predictors are easily obtained features from forum facilities, such as upvotes, unique views, and unique collaborations. For ground truth, we used expert judgment by current teaching assistants for the same course.