ZhihuRank: A topic-sensitive expert finding algorithm in community question answering websites

Abstract

Expert finding is important to the development of community question answering websites and e-learning. In this study, we propose a topic-sensitive probabilistic model to estimate the user authority ranking for each question, which is based on the link analysis technique and topical similarities between users and questions. Most of the existing approaches focus on the user relationship only. Compared to the existing approaches, our method is more effective because we consider the link structure and the topical similarity simultaneously. We use the realworld data set from Zhihu (a famous CQA website in China) to conduct experiments. Experimental results show that our algorithm outperforms other algorithms in the user authority ranking.