An approach to rapid worker discovery in software crowdsourcing

Abstract

Crowdsourcing is an emerging business model which organizes distributed crowds to solve various problems through the Internet. Recently this paradigm has also flourished in software engineering domain. Despite benefits like on-demand workforce and rich expertise, a major challenge is how to discover cost-effective workers while guarantee high quality of deliveries. Current approaches to worker discovery have primarily based on workers’ interest and self-evaluation which have subjective deviations. In this paper, we present a novel approach to describe each skill on the basis of feedback evaluation and estimate skill evolution dynamically. On the other side, to deal with the growing number of workforce, we use a clustering-based method to group workers with similar characteristics together to reduce search space. Our experiments show the effectiveness and efficiency of this approach.