Moving from static to dynamic modeling of expertise for question routing in CQA sites

Abstract

CQA sites are dynamic environments where new users join constantly, or the activity levels or interest of existing users change over time. Classic expertise estimation approaches which were mostly developed for static datasets, cannot effectively model changing expertise and interest levels in these sites. This paper proposes how available temporal information in CQA sites can be used to make these existing approaches more effective for expertise related applications like question routing. Adapting two widely used expert finding approaches for question routing returned consistent and statistically significant improvements over the original approaches, which shows the effectiveness of the proposed temporal modeling.