Tech & Check: automating fact-checking
Part of a series of interviews with organizations that are funded via the Knight Media, Trust and Democracy Initiative. We begin with Bill Adair, director of Duke Reporters Tech & Check project.
NW: What’s your elevator pitch on “Tech and Check?”
BA: We are helping fact-checkers by automating some of their most mundane tasks and using automation to disseminate their work to much larger audiences.
NW: How do you define mundane tasks?
BA: I had a conversation once with Chengkai Li, a computer scientist at University of Texas, in which he asked me, “What do fact-checkers need?” And I said, “Fact-checkers need interns.” At any fact-checking organization it’s usually the interns who go through transcripts and op-eds and identify factual claims to be checked.
The process can be time consuming and pretty mundane, so Chengkai came up with an algorithm that can predict when a sentence is a factual claim that a fact-checker would like to check.
Our Claimbuster products are taking what he built and creating automated ways to submit content to his API and then push the results out to fact checkers. I think of our tools like interns who watch cable TV or sit in the House gallery but never get tired.
NW: What milestones have you reached with the project so far?
BA: Our script is analyzing cable news transcripts and then auto-generating an email every morning that contains claims they may want to check. We’ll be sending these to fact-checkers soon.
NW: What do robots do wrong and what do they do really well?
BA: On the Claimbuster products, we’ve had to teach the robots the difference between a TV news reporter, and a pundit or politician. Another challenge is the general messiness of how transcripts are presented. As with any sort of human entered data, transcripts are typed in by people and those people are not always consistent.
NW: Tell me about what you are doing to bring fact-checking to broader audiences.
BA: We’re developing a series of products that will pop up fact-checks at the moment the consumer needs them the most, when they hear the factual claim. Our first product will be an iPhone app we hope to test during the State of the Union address on January 30. That app will allow fact-checkers to push notifications in real time. In the future we’ll explore how to do these pop-up fact checks on a TV using Apple TV or Chromecast. Eventually, we plan to fully automate it, as we’ve done with our Share the Facts app for the Amazon Echo.