The U.S. Senate's long resistance to filing campaign finance reports on anything but paper occasionally turns into a hot political topic: At the moment, for example, Republicans are complaining that Democrats are intentionally swamping the system by submitting minutely-detailed reports to the Federal Election Commission.
By the next election cycle, it might not matter: Smart technology is being put in place to make Senate candidates' finances public in near real-time.
California-based Captricity has just been awarded a $270,000, one-year contract to quickly extract all that data trapped on paper and push it onto the FEC Web site in, potentially, hours. How?
"We cheat," says Kuang Chen, Captricity's chief executive. "We use crowd-sourcing and machine-learning." The 26-employee company grew out of Chen's 2011 PhD thesis in computer science at the University of California at Berkeley.
Captricity works by "shredding" documents into small pieces, explains Chen. Those pieces are then uploaded to Mechanical Turk, Amazon.com's distributed workforce platform, where they're farmed out to people all over the planet. (Amazon chief executive Jeffrey Bezos owns The Washington Post.)
Each Turk worker examines the sliver of document before them and identifies the few characters on it -- such as the number "5" or letter "S," for example. Their answers are used to train a computer algorithm that the company is developing for the FEC to anticipate what long strings of text are likely to appear in certain fields, such as "$5-0-0" or "S-a-n D-i-e-g-o." The software gets smarter each time it runs, says Chen, with its accuracy "asymptotically approaching 100 percent."
Captricity has done work with California's election commission, and there it caught the attention of Ann Ravel, the vice chair of the Federal Election Commission, who was then chair of the California Fair Political Practices Commission. "When I got to the FEC in 2013," Ravel says, "I realized the antiquated way they were doing Senate filings."
Candidates who opt not to submit their fundraising data electronically file to the Senate Public Records Office on paper; the Senate administrators scan them in as images, and those files are then passed along to the FEC. The FEC takes a copy of the image and converts it into a PDF and posts it online. The PDFs are then sent along to a contractor, whose workforce manually reenters the documents on a computer. The resulting electronic files are them posted to the election commission's Web site at FEC.gov.
That slow process costs the FEC about half a million dollars a year. "It's crazy," says Ravel.
And occasionally, it's simply too much. As of Friday afternoon, the FEC Web site featured a scrolling notice admitting that it just couldn't handle all that paper:
Unusually large paper reports filed by U.S. Senate candidates in the third quarter of the current cycle have overwhelmed our processing capacity, slowing public disclosure of those reports. We regret the delay and are taking urgent action to publish copies of all Senate reports as quickly as possible.
Soon enough, they might be able to retire that banner.