Authority vs Affinity: Modeling User Intent in Expert Finding

This paper considers the problem of recommending experts for a question. The problem of expert finding has been investigated in the past, within the context of information retrieval, where the focus has been on measuring expert authority. However, we treat the problem as a recommendation problem in the context of social search, that is, expert recommendations are personalized for a questioner based on his/her history. The intuition is that questioners look not only for authoritative responders, but for a combination of authority and personal trust or affinity. The significance of these factors depends on user intent in asking the question, and varies with each question, and with questioner personality. An iterative algorithm is proposed to model user intent for a question in terms of whether they are looking for authoritative experts, or individuals in their social networks, and an expert recommendation algorithm based on it is tested on data crawled from the Yahoo! Answers website.