MATH Seminar

Title: Harnessing the Power of Crowd for On-Demand Geographical Data Collection
Colloquium: Computer Science
Speaker: Cyrus Shahabi of University of Southern California
Contact: Li Xiong, lxiong@emory.edu
Date: 2014-02-20 at 12:00PM
Venue: MSC W303
Download Flyer
Abstract: GeoCrowd is an online spatial crowdsourcing market (similar to Amazon’s Mechanical Turk) that matches geo tasks (i.e., tasks associated with a specific location and time such as “Take pictures of Tommy Trojan during 2012 USC-UCLA game”) to human workers. Every person with mobile devices can now act as a multi-modal sensor collecting various types of data instantaneously (e.g., picture, video). With GeoCrowd, subscribers can publish tasks with specific space and time attributes. Subsequently, the workers (with GeoCrowd mobile app) can perform the tasks if they are at the right time and at the right place and upload the results to the GeoCrowd server(s). In this talk, I first introduce our generic framework for GeoCrowd and discuss various techniques for optimal assignment of spatiotemporal tasks to human workers. Next, I show how we can extend this framework to incorporate trust in GeoCrowd in order to ensure workers satisfy a confidence value given by the task requester. Finally, I will show an application of the GeoCrowd framework in a commercial domain.