How Amazon Teamed Up With Twitter On Its Human Search Army

Amazon box with Twitter bird superimposed on top. Edited by TPM. Original image by Mike Seyfang, posted to Flickr with CC-by-SA 2.0 license (share alike).
Start your day with TPM.
Sign up for the Morning Memo newsletter

Twitter earlier this week revealed a surprising fact: It uses Amazon’s Mechanical Turk, an online labor marketplace where companies can farm out low-paying, low-skilled jobs to willing workers, to find the answers to Twitter search queries.

Amazon came forward on Thursday to elaborate on its partnership with Twitter and how it hopes other companies will use the labor marketplace.

“Mechanical Turk provided Twitter with the tools to quickly develop their custom workforce,” wrote Rena Lunak, a spokesperson for Amazon Web Services, in an email to TPM.

The two have been working together “for a while,” according to Amazon, prior to their appearances together last year at the Social Media and Web Analytics conference and the CrowdConf crowdsourcing conference.

Asked specifically how Amazon and Twitter collaborated, Lunak said its Mechanical Turk team received feedback from Twitter on “how we can help them be successful.”

“In terms of the development of the project, the teams from Twitter and Mechanical Turk had discussions as it was being developed and this is common with many of our Requesters,” Lunak added.

Amazon considers companies that post jobs on Mechanical Turk to be “Requesters,” and the jobs themselves to be “human intelligence tasks,” or HITs, though many involve simple, quick tasks such as looking up contact information online or generating adult content titles based on provided images.

Twitter in a blog post Tuesday stated that the company created its own group of custom workers, “culled from the best of Mechanical Turk,” to ensure reliability and accuracy in the quality of the search query answers they provided.

But when asked by TPM exactly how Twitter went about identifying and selecting its custom workers, or “Turkers,” and how many Twitter relied upon, Twitter declined to specify.

Amazon’s Lunak shed slightly more light on how Twitter was able to build its Amazon Turk dream team:

“Mechanical Turk provides a number of methods for Requesters to control who can complete their work. Twitter uses one of these methods in which they create a qualified group of Workers who can perform their work.”

Additionally, another built-in Amazon Mechanical Turk feature seems to fit Twitter’s needs perfectly.

“Requesters can use Master Workers (an elite group of Workers who have demonstrated high accuracy across many Requesters in the Mechanical Turk marketplace) or use generic qualifications such as a Worker’s approval rate or location,” Lunak explained.

However, Twitter in its blog post said it did not “use the standard worker filters that Amazon provides.” The company said it had “experimented with these” but concluded “a custom pool fit our needs best for a few reasons…”

Overall, Amazon thinks Twitter is “a great example of a company that is successfully leveraging the Mechanical Turk marketplace for improving real-time search with human evaluation,” according to Lunak. However, Lunak was quick to point out that Twitter was but one of thousands of businesses using the service, which launched in 2005.

“The Twitter use case is similar to use cases we see from many of our Requesters,” Lunak said. “Businesses – from startups to enterprises — are using Mechanical Turk for a broad range of use cases, including moderating user generated content, data cleansing, transcription and information gathering and research.”

Amazon’s promotion of Mechanical Turk as a tool for other Web brands makes sense given the fact that it takes a cut10 percent — of all payouts by companies to workers. But it makes sense in a longer tail play as well: Mechanical Turk is part of Amazon’s Web Services, the cloud-based products and platforms Amazon provides to many other large Internet companies. Amazon’s elastic cloud server (EC2), which hosts heavily trafficked websites including Netflix, Instagram, Pinterest and Reddit, among many others, is one primary example.

Although the company may compete with some of those websites for eyeballs, it also benefits from their dependence on Amazon’s architecture for their own business. Amazon’s support for Twitter Search through Mechanical Turk seems to be based on a similar premise.

Latest Idealab
Comments
Masthead Masthead
Founder & Editor-in-Chief:
Executive Editor:
Managing Editor:
Associate Editor:
Editor at Large:
General Counsel:
Publisher:
Head of Product:
Director of Technology:
Associate Publisher:
Front End Developer:
Senior Designer: