Identifying Potential Experts on Stack Overflow

Abstract

Question answering community is an online service of user-generated content, where users seek help by posting questions and help others by offering answers. In question answering community, most of high quality answers are posted by some users called experts. The early identification of experts is of great significance to the success of community, based on which we can take measures to avoid the loss of expert users and encourage them to make more contributions. Different from the related works, we put forward an efficient method of supervised learning to identify potential topical experts in question answering community. Above all, we define and quantify the concepts of expert. Then on a specific topic, we extract the user features from three dimensions, including text-feature, behavior-feature and time-feature. Finally, we use the classification algorithms in machine learning to identify whether a user is the potential expert on current topic. Based on the data of Stack Overflow, we carry out a lot of experiments and implement a potential experts identification system. The results demonstrate the excellent effectiveness of our method based on artificial neural network model. Besides, we find that expert users are inclined to interact with other expert users, providing new ideas for future research on this subject.