Robot Reporters: Software Turns Raw Data Into Sports, Financial Reports
ROBERT SIEGEL, HOST:
So computers are deciding whom to hire, which is increasingly themselves.
AUDIE CORNISH, HOST:
The robot takeover has begun. All right, we won't go all "War Of The Worlds" on you here. We're journalists.
SIEGEL: But we do have competition. There's software out there that generates news reports from raw data. We're talking about algorithms in this case, too.
KRIS HAMMOND: It'll grind through what it has to do and produce something that is absolutely human-readable and, in fact, is indistinguishable from something a human has written.
CORNISH: That's Kris Hammond with a company called Narrative Science. Hammond developed software called Quill. It can take data - say, sports scores - and turn the numbers into stories with a beginning, middle and end.
SIEGEL: These aren't reports about March Madness. Think Division III baseball, high school football, Little League.
HAMMOND: So parents and coaches can score Little League games. And you get the numbers, but that's all you get.
SIEGEL: A human reporter may not take the time to write up a story of little Timmy's T-ball homerun, but it's no problem for an algorithm.
HAMMOND: Who's going to read those? Well, maybe 10 kids for any given story. But for those kids, it's really valuable.
SIEGEL: Which is the main benefit of automation - giving the computers all the work that we don't want to do.
CORNISH: So how does all this work in an actual newsroom? We asked Lou Ferrara.
LOU FERRARA: I get asked about this a lot - automation technologies - because people think I have robots in the newsroom and, you know, we have, like, "Terminator" type settings here or something.
CORNISH: He's a vice president at the AP.
FERRARA: I oversee sports, business and entertainment news, as well as interactive and social and a couple other cool things at The Associated Press.
CORNISH: Cool things like their automated news program, which started churning out local sport news and financial reports over the Associated Press wire. It's similar to the program that generates Little League stories.
SIEGEL: The AP software looks out for earnings data posted by companies and then uses an algorithm to plug it into a story template.
FERRARA: So if a company earned X above Y in a certain quarter, then it gets this story. If it had a loss that quarter, it gets another version of the story and so on.
CORNISH: Ferrara says this frees up his human reporters to work on other things. Meanwhile - the software? It's fast.
FERRARA: And so within seconds or minutes really from an earnings report's being announced by a company, we have it out on the AP wire.
SIEGEL: But, Audie, don't start to worry about a "Robocop" world just yet.
(SOUNDBITE OF FILM, "ROBOCOP")
UNIDENTIFIED ACTOR: This guy is really good.
UNIDENTIFIED ACTOR #2: He's not a guy. He's a machine.
UNIDENTIFIED ACTOR: What are they going to do, replace us?
SIEGEL: They're still a long way from launching the radio host automaton.
CORNISH: Whatever you say, robot - I mean, Robert.
NPR transcripts are created on a rush deadline by Verb8tm, Inc., an NPR contractor, and produced using a proprietary transcription process developed with NPR. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.