When it comes to Big Data, machines can only do so much. At least that’s what startup Spare5 Inc. is betting with the launch of what it calls an “Intelligent Crowdsourcing Platform” that leverages a community of specialists to process Big Data tasks that require a human touch.

Crowdsourcing services today run the gamut from the mundane (Amazon.com’s Mechanical Turk) to the highly specialized (CrowdMed). Spare5 sits somewhere in the middle, but its special sauce is a machine learning algorithm that rates the skills and preferences of crowd members.

The platform is useful for “pretty much anything that involves getting structured data out of unstructured data,” said Matt Bencke, founder and CEO. Examples including adding keywords and categories to photographs or looking up information online that isn’t easily found by keyword search.

The service uses a gaming metaphor to lighten the tasks and it breaks up jobs into small components that can be handled quickly on a smart phone. “It’s strictly data or questions you can complete wherever you are,” Bencke said.

Crowd members, whom Spare5 calls “Fives,” are paid a few cents for each task. Anyone can sign up to be a Five. As tasks are completed, Spare5’s machine learning engine assesses the skills and interests of participants and assigns them more specialized jobs.

“As each new answer comes, in we score whether they’re good or bad at a task,” Bencke said. “So they might be good at copy editing but not at sports. We line them up with tasks that are right for them.” The company has recruited more than 40,000 Fives.

Customer Avvo Inc. has been using the service to assess and categorize a library of more than 40,000 legal guides that it maintains on its website. Avvo is itself a form of crowdsourcing. It matches consumers’ legal questions with qualified lawyers who provide answers for a fixed fee. The site attracts about 8 million unique visitors per month.

The legal guides are an essential part of Avvo’s value proposition because they help people with legal questions find the site and drive repeat visits. “Quality is a huge factor, and we want to be able to provide the best information at scale,” said David Holmberg, an Avvo product manager. The company was unable to develop algorithms that could reliably assess quality, so it turned to Spare5 to find people with enough legal knowledge to compare and rate the documents.

“Typical crowdsourcing isn’t designed for that level of sophistication,” Holmberg said. “It’s designed for high-volume and simple tasks.”

Spare5 now processes about 1,000 documents for Avvo each day. Based on audit results, Avvo figures Spare5 achieves about a 94 percent accuracy rate in correctly assessing the quality of legal guides, Holmberg said. “We looked at a number of different crowdsourcing services, but the level of sophistication Spare5 was able to offer at scale was so much better,” he said. He declined to reveal how much Avvo pays for the service, but the company’s director of legal marketing said Avvo expects a 60 percent increase in onboarding of new customers this year as a result.

Spare5 has raised a little over $13 million since its founding in 2004.

CC image via Wikimedia Commons