Sports writing reduced to a science
- Share by Email
The men pictured above (l. to r.) are Kris Hammond, Larry Birnbaum, and Stuart Frankel of Narrative Science, and their hard work is a potential haymaker to the jaws of sports journalists everywhere. They are the evil geniuses responsible for a sportswriting machine profiled in yesterday’s New York Times. And I ain’t even mad.
Their computer program wrote a game brief within 60 seconds for a Big Ten website based on data, developed an algorithm (or whatever) for more human-like writing, and amazed some experts who don’t really know quality sports writing, but are experts nonetheless.
“I thought it was magic,” says Roger Lee, a general partner of Battery Ventures, which led a $6 million investment in the company earlier this year. “It’s as if a human wrote it.”
Experts in artificial intelligence and language are also impressed, if less enthralled. Oren Etzioni, a computer scientist at the University of Washington, says, “The quality of the narrative produced was quite good,” as if written by a human, if not an accomplished wordsmith.
On the bright side, that blockquote contains optimism from biased parties. An enthusiastic investor should marvel, and a high compliment from a scientist is cool, but can he write a 400-word gamer? The computer-powered story Steve Lohr used as an example in his lead is acceptable, but nothing truly remarkable. Now, if only Narrative Science’s supporters can convince the scientists to add more hyperbole to its prose.
The machine’s writing is precise and accurate. It delivers news as information without feeling, pumping out the kind of content human writers often prefer to publish without a byline.
Sports writing programs aren’t a new concept, likely because most sports writing feels programmed on its own. Game capsules for previews and recaps are often a regurgitation of statistics and meaningless history. Such articles usually read as if there’s no human element to the writing (or the game), anyway. A lot of it is pure information in an industry suffering from information overload.
The good thing is that Narrative Science’s work, thus far, isn’t exactly traffic-driving content. It’s a recap people will only see if linked on Twitter, or if they noticed it while already browsing the site. The publisher’s SEO practices likely matter most.
But if Narrative Science teaches the machine to do WAG slideshows, the whole business is in trouble.
Business Week wrote about these guys a year and a half ago, asking if sportswriters were really necessary. It’s a dick question, but one that’s given credence when most people think sports writers are merely ex-jocks living vicariously through actually talented athletes.
Nonetheless, the idea has been circulating for more than a decade, but was never executed well enough to be considered a legitimate option. Narrative Science believes they’re close to such legitimacy. I’m reserving judgment until the software can attend games and interview players. Human writers might have a place until then.
I’ll say the sky is falling once the program proves savvy enough to aggregate locker room quotes as they appear on Twitter.
As students, we’re trained to write; and then as writers, we watch other writers attain various levels of success while middle-fingering the rules we were taught. The frustration sets in for those of us who straddle this line of compliance while stifling our instincts, only to fall behind and watch those rules get exploited by a computer that computes.
When I started writing, professors and editors nixed entire graphs with a reminder that there’s no room for my opinion, that such words belong in a column and not a straight-news piece. Now, it’s becoming more apparent that there’s actually no room for lengthy, straight-news stories.
I edited Associated Press reports as part of my intern duties in the past, condensing full stories into accurate 40- to 60-word blips, and learned something very important: Pure recaps suck. And they have for a very long time.
There are only so many ways to write ‘this team beat that team because that guy did this.’ The writing was robotic and formulaic way before Narrative Science programmed a motherboard to handle the task. I Tumblr’d about it earlier. Allow me to quote myself:
… this is also an indictment on sports writing and how easily these bland recaps are yawned out.
No one cares about the recap anymore. We have 24-hour highlight shows, YouTube, and apps for that shit — with 140-character Twitter updates to churn out stat lines and snark. And then fans don’t have to care anymore, because the next morning is for thinking about the next game.
The recap has needed more imagination for a long time, because readers don’t want to know what happened as much as why it happened.
I have no choice but to respect Narrative Science, even if they want to Alt+Ctrl+Delete me. I’ve wanted to delete and nuke a lot of the bland sports writing and reporting that’s already killing the field.
Hammond is looking for a Pulitzer from his program within five years. Writers need to think about what we’re going to do before then.
“In five years, a computer program will win a Pulitzer Prize — and I’ll be damned if it’s not our technology.”
It’s time for journalism to evolve and I’m sure of it. There are no definitive answers to the how, but I’m convinced it starts with changing the narrative and twisting up the scientific process.