An Evaluation Algorithm for Expert in Community Question Answering by Combining Topics and Behaviors

Abstract

In this paper, we propose an algorithm to evaluate the influence of experts in community question answering (CQA). In the current research of influence evaluation algorithm, the main research method is to establish the network between users and problems, and use centrality algorithm such as PageRank to evaluate the influence of users. However, Existing algorithms tend to favor users with more followers. In this paper, we propose an evaluation algorithm based on user’s content combined with link relationship. The experimental results show that our method can give higher ratings to users who have low number of followers but can provide high-quality content. At the same time, we add the topic relevance to the evaluation algorithm, so that the expert score will change according to the different topics.