The National Science Foundation (NSF) has awarded a University of Texas at Austin School of Information faculty member $550,000 to explore ways to make crowdsourcing more viable and lower the risk for potential adopters.
The grant is one of three early career awards that assistant professor Matt Lease has won in the past year.
"The challenge of quality assurance continues to deter many employers and institutions from applying crowdsourcing practices in their organizations and reaping the benefits," said Lease, who is a pioneer in the nascent field of crowdsourcing research. "For those unfamiliar with the term 'crowdsourcing,' it simply refers to the practice of mobilizing a large, global population to complete tasks or offer information and data."
The NSF award will allow Lease to investigate, integrate and benchmark different quality assurance algorithms across a wide range of tasks, dataset sizes, labor sources and operational settings.
"The goal of the work is to develop a practical, comprehensive set of best practices for potential and current crowdsourcing users," said Lease. "Investigation of techniques to enable massive, real-world crowdsourcing datasets will push the scale an order of magnitude beyond what researchers commonly study today."
In addition to the NSF's CAREER Award, Lease also has received $290,000 from the Institute of Museum and Library Services (IMLS) to investigate cost-effective curation practices that will make it easier to preserve everyday conversations and allow them to be added to our cultural record. Complementing this IMLS project, he also received a $300,000 Young Faculty Award from the Defense Advanced Research Projects Agency (DARPA) to apply enhanced speech transcription technology to improve state-of-the-art search engine technology for searching conversational speech archives.
"Just as pen and paper once did for writing, technology now allows us to capture and preserve our conversations for posterity," said Lease. "Unfortunately, storing all this data won't be very useful unless we can effectively search it."
The complementary IMLS and DARPA projects will investigate ways to make it as easy to search past spoken conversations as it is to search old emails.