Crowdsourced data preprocessing with R and Amazon Mechanical Turk

Abstract

This article introduces the use of the Amazon Mechanical Turk (MTurk) crowdsourcing platform as a resource for R users to leverage crowdsourced human intelligence for preprocessing "messy" data into a form easily analyzed within R. The article first describes MTurk and the MTurkR package, then outlines how to use MTurkR to gather and manage crowdsourced data with MTurk using some of the package's core functionality. Potential applications of MTurkR include construction of manually coded training sets, human transcription and translation, manual data scraping from scanned documents, content analysis, image classification, and the completion of online survey questionnaires, among others. As an example of massive data preprocessing, the article describes an image rating task involving 225 crowdsourced workers and more than 5500 images using just three MTurkR function calls.