On the value of answerers in early detection of response time to questions for peer recommender systems

Abstract

Most research in peer recommender systems in online learning communities (OLCs) is focused on the problem of identifying the answerers who can provide the best answers to a question soon after the question has been asked. In fact, we have explored exactly this issue in another research project [5]. In contrast, in the research reported in this paper we devised methods of predicting in Stack Overflow (a very large online community of programmers) at the time a question is asked, whether the question will receive an answer at all, and, if so, whether the answer will come early or late (i.e. after the median response time). Overall, in a study that used 5 months of Stack Overflow data our methods worked well enough that we feel they could usefully inform support systems such as peer recommender systems about questions that might prove to be unanswered or answered late.