Instant expert hunting: building an answerer recommender system for a large scale Q&A website

Abstract

After years of rapid growth, Q&A websites, such as Yahoo Answers and Baidu Zhidao now start to utilize recommender systems to route questions to potential answerers. Traditional studies on question recommendation on Q&A website usually presume that answerers are always instantly available for askers. They neglect the importance of timely response of answerers to question askers. Such negligence of time-efficiency leads to lower quality of recommender systems, namely the question starvation problem. In this paper, we focus on improving time-efficiency of recommendations on a large scale online Q&A website in order to boost the answer ratio of new posted questions. To achieve this goal, we design an answerer recommender system (ARS) that is capable of instantly routing a new question the potential answerers with the right expertise. We test our system on a real-world large-scale Q&A website. Experimental results demonstrate that compared with a previously used baseline system, in our system both the time-efficiency and the number of answers given to a question are dramatically improved. In addition, we find that timely response to questions can stimulate interactions between question askers and answerers. Such interaction can boost the answer acceptance ratio by askers since it encourages answerers to improve their answers actively and continuously.