January 29, 2013

Truth Teller | Knight Foundation
Politicians lie.

Journalists try to point out those lies, but usually at some later time and in a different medium. That gap in time and distance is just enough to let the original lie take root in the public’s mind before the truth catches up, if it ever does.

Closing that gap is the holy grail of professional fact checkers. PolitiFact made some progress in last year’s election with the Settle It! mobile app that empowers users to look up fact-checks at the precise moment they need them. Dan Schultz at MIT has worked on a browser plugin called Truth Goggles that highlights truths and falsehoods on whatever Web page you’re currently reading.

Today, The Washington Post introduces a new prototype, The Truth Teller, that does live, automated fact checking of a political speech. It leans on some of Schultz’s technology and also cites PolitiFact and Factcheck.org content in addition to the Post’s own fact-checks.

Specifically, it attempts to figure out what a person is saying, extract distinct factual claims from his speech, and link to pre-existing fact checks of similar claims.

In one of the prototype videos, House Speaker John Boehner makes a false claim that raising taxes on the wealthiest Americans would destroy 700,000 jobs.

Washington Post Executive Producer for Digital News Cory Haik explains in a blog post how it works:

The Truth Teller prototype was built and runs with a combination of several technologies – some new, some very familiar. We’ve combined video and audio extraction with a speech-to-text technology to search a database of facts and fact checks. The Post also worked with Dan Schultz, creator of Truth Goggles, as he helped consult and shared his knowledge of real-time fact checking. We are effectively taking in video, converting the audio to text, matching that text to our database, and then displaying, in real time, what’s true and what’s false. The key to the project’s success is building an authoritative database – our goal is to identify falsehoods, not create more of them.

The project is funded by a $50,000 grant from the Knight Foundation’s Prototype Fund. The idea came from Post national political editor Steven Ginsberg, who told Craig Silverman last summer he was inspired while listening to a politician in Iowa.

“It was one of those small parking lot affairs outside a sports bar and the candidate was there speaking to about 30 people … For about 45 minutes he said a lot of things that I knew to not be true, and nobody else there knew that.”

Ginsberg called Haik and the seed was planted.

In a phone interview Tuesday, Haik told me the goal for the near future is to get the technology reliable enough to integrate with the Post’s livestreaming platform, so fact-checking data pops up in the same place and time that viewers are watching an event.

That’s going to require that the algorithm continue to improve so it consistently and accurately assesses whether spoken claims are true or false, Haik said. It also will require the Post to build out a larger database of fact-checks and to develop a structured data system of organizing them.

In the long run (perhaps for the 2016 election cycle) Haik envisions plugging this into a mobile app.

“Hold your phone up in the middle of a field in Iowa when Michele Bachmann is speaking, record that audio from your phone and have that play through and tell you whether it’s the truth or not — that’s still our goal,” Haik said. “For people to be able to gather audio and video in places where politicians are speaking that we might not be streaming or have access to.”

Support high-integrity, independent journalism that serves democracy. Make a gift to Poynter today. The Poynter Institute is a nonpartisan, nonprofit organization, and your gift helps us make good journalism better.
Jeff Sonderman (jsonderman@poynter.org) is the Digital Media Fellow at The Poynter Institute. He focuses on innovations and strategies for mobile platforms and social media in…
Jeff Sonderman

More News

Back to News


Comments are closed.