Expertise ranking in question-answer social network groups

Online social networks have become the major channel for people to maintain relationships, collaborate, and contribute shared information. A social network group can be created as a specific community for people who share an interest in a particular topic. One form of interaction within the group is the question-answer interaction by which the users in the group can ask questions or provide answers to others. A network analysis method, e.g., PageRank, can be used to analyze the interaction patterns between the users in order to identify and rank experts in the group. In this paper, we are interested in experimenting on how the quality of the users' comments can take part in the identification and ranking of experts by a PageRank-like algorithm. The quality factors are community rating, that is given to the answers, and the content-based features of the comments, i.e., length, complexity, and informativeness. We conduct an experiment on a Java Facebook group and evaluate the accuracy of the ranking with and without comment quality consideration against expertise ranking by Java experts.