User profiling for answer quality assessment in Q&A communities

Collaborative Web applications, such as forums and question answering websites, help users to get answers to a wide array of questions. Given the large amount of information available, it is important to devise automatic methods that surface high quality answers. Our objective here is to determine if there are any particular aspects of a user's profile or activity in the community that can be exploited to spot high quality contributions. We first perform an in-depth analysis of the information provided by the users in their profiles in order to discriminate features that are correlated to expertise. Then, we propose an answer ranking scenario in which we assess the predictive capabilities of profile and activity related features. In our experiments, we use a large scale corpus from Stackoverflow, a very active Q&A community focused on technical topics and find that answer rankings obtained using a user model outperform a ranking based on the chronological order of answers.