A Divide-and-Conquer Approach for Crowdsourced Data Enumeration

Abstract

Crowdsourced data enumeration, in which the Web crowd is requested to enumerate data items within a specified range, is important in many Web applications such as hotel reviews. This paper presents a processing method for crowdsourced data enumeration on microtask-based crowdsourcing platforms. A general approach to achieving a high recall in data enumeration is to apply the divide-and-conquer principle. However, how to apply the principle to data enumeration on microtask-based crowdsourcing platforms is not trivial. The proposed method is unique in that the workers join the process of generating smaller tasks in a divide-and-conquer fashion, and the programmer does not need to provide many microtasks in advance. This paper explains the method, provides theoretical results to show the method works well with microtask-based platforms, and explains our experimental results that suggest the proposed method can achieve higher recalls and produces appropriate tasks for microtask-based crowdsourcing.