Amazon.com's Web services subsidiary released a version of its Mechanical Turk on-line work force service that
offers a new GUI for developers and business analysts. The goal is to ease the programmability of Web service-based workflow applications.
Amazon Web Services LLC's Mechanical Turk is an online marketplace that supplements largely machine-centric Web 2.0 workflows with human helpers. It now offers a more intuitive set of Web-based tools, according to Sharon Chiarella, vice president of Amazon Mechanical Turk. The tools are designed to make it easier to add the Turk's Human Intelligence Tasks (HITs) to business processes within the enterprise, she explained.
While the interface for the humans who signed up to do work was graphical, on the enterprise side it required coding to integrate Mechanical Turk functionality into business processes. The new set of GUI tools is designed to allow business analysts and even non-technical business users to access the human task marketplace into their Web applications, Chiarella said.
The Amazon original version, launched in November 2005, introduced a set of a Web services APIs that were advanced for the time. A need to move beyond command line interfaces and WSDLs is at the heart of the new version.
The Amazon Mechanical Turk takes its name from a pre-Industrial Age chess-playing automaton. The 'machine' was said to have beaten Benjamin Franklin in a game of chess. In fact, that machine's success was due to a simple trick: the 'CPU' was a small human being with exceptional chess playing attributes hidden within the mechanism.
The Amazon 'Turk' software is related to the original in name only, as humans are openly integrated with Web services to create what Chiarella describes as an online "marketplace for work that requires human intelligence." People sign up to work online doing tasks that even state-of-the-art artificial intelligence cannot handle, such as ranking the impact of digital photos, identifying performers of digital music, or re-writing corporate documents, she said.
"The applications really vary," Chiarella said. "We see a number of companies using it for algorithm refinement. For example, if you have a search algorithm and you want to verify the relevancy of the results you are getting back that's a great application for Mechanical Turk."
In that case a company might apply human intelligence to verify that a search term such as "digital camera" is returning results that match the business requirements, she said.
"That helps developers see how well they are doing and fine tune their algorithm," she said.
Unlike a traditional in-house quality assurance department where perhaps two testers might be asked to check the algorithm results, Mechanical Turk allows a company to temporarily increase the number of workers doing the testing, Chiarella said.
"You can ask a hundred workers, or 50 workers or 25 workers to get a sense of how the search results play to a broader audience," she said.
For Web 2.0 governance, Mechanical Turk can be used for policy enforcement that goes beyond the capabilities of WS-Policy based programs.
"A lot of Websites allow users to add comments to blogs or actually create content," Chiarella said. "Those Websites have a set of guidelines. Rather than hire a team of editors to keep on top of all the entries, you can ask the [Mechanical Turk] workers to check for appropriate language. Or if it is a coherent posting."
In this way, Mechanical Turk can help organizations provide policy enforcement for Web 2.0 applications as "content explodes," the Amazon vice president said.
Amazon's business model for this Web services offering is based on getting a 10 percent commission on the fee that businesses agree to pay the humans working under the covers of the Mechanical Turk, Chiarella explained.