By Malik Singleton, Developer, NationTimes.net
Nowadays cars park themselves, remote SWAT bots snatch perps, and there are even robo-bartenders. So, inevitably, journalists are also having to share the stage with robots — case in point: here is a 1705-word story with no byline. Why? It could very well be because a robot wrote it.
Recently, I attended a Meetup and learned about a technology called Narrative Science, a service that “turns data into stories.” This service, already being used by a few newsrooms, processes statistics then associates those numbers with pre-programmed rules of grammar and phrasing to ultimately spit out a “story” that is readable and natural-sounding to humans.
The best examples of its results are with number-driven stories such as baseball recaps, financial earnings reports and real estate analyses. The Narrative Science engine actually tries to figure out what the numbers mean. It interprets highs and lows, compares current performance with historical averages, intelligently evaluates trends or anomalies, and produces prose that easily amounts to a kind of reporting, but the question for real people is: is that really journalism?
Larry Adams, a real person at Narrative Science, says, “it absolutely is journalism.” The VP Product Management defends the authenticity of the service’s reporting because he says there are living and breathing writers behind everything, though in less traditional roles.
“We have editorial staff from Medill and Columbia training the system. The content has the voice of the writers and gets produced via automation.” So in the case of earnings reports, Adams says they read “tons” of reports to deconstruct the angles that journalists use.
“There are only so many hundreds of combinations of angles, so we establish rules.” They call these rules representative language and they assign what they call “interestingness” to certain types of information to teach the system qualitative news judgment. We still ask the question whether this is all we human news people do or if we do something more.
Can a computer really connect the dots like a thinking person can? Don’t computers lack the sensibilities to find the emotional elements of a story? Rich Gordon of MediaShift’s Idea Lab says, “human journalists will do — and should do — the kind of reporting and storytelling that computers can’t.”
Adams says human journalists should, in fact, focus more on interviews and less on tasks like research. “Narrative Science is not going to be able to interpret body language, but it employs a vast history of information with a breadth of coverage that an entire human staff can’t do.”
This thinking is what has led some newsrooms, including the Big Ten Network, to adopt Narrative Science. The startup company is betting that news organizations will consider that it is too expensive and logistically impossible to have reporters cover every newsworthy story. Not to mention the chore of certain types of coverage, some real reporters actually get bored with writing mundane reports — so why not switch their focus to things like one-on-one interviews, enterprise stories, human interest pieces, in-depth investigations, etc.?
Although Narrative Science is still being fine tuned and only works for a few types of stories, the big impact it’s having so far is that it’s forcing us to have a very critical debate.