摘要
Compared with community-based question answering systems and modern search engines, social network-based question answering systems are more efficient in addressing non-factual questions. In such systems, askers search answerers among their 1-hop neighbors; however, high-quality answerers may exist in the k-hop neighbors of the social networks who are not known to askers directly. To address this problem, we propose a dynamic SOcial network-based Question Answering System (SOQAS) that finds high-quality answerers to each asker's question with high response rate and low-response time. The SOQAS finds high-quality answerers in the k-hop dynamic social network and selects optimal relays at each hop to forward the question to, via social referral chains. In particular, the profile information is exchanged among k-hop neighbors, and leveraged for finding high-quality answerers and optimal relays at each hop, so as to increase the response rate and reduce the response time. We conduct trace-driven simulations, which show that, compared with the state-of-the-art schemes, SOQAS achieves: 1) higher average expertise levels by more than 42%, 2) higher average response rate by more than 26%, and 3) lower response time with as high as 27% reduction. Furthermore, under diverse system parameters, such as question arrival rate, keywords per question, answerers per question, number of hops, and predictability, the SOQAS consistently outperforms the state-of-the-art schemes.