But it doesn’t stop there: Computer systems can now write the stories and design the advertisements that surround them.
The NAA mediaXchange in Washington in April brought a handful of such smart systems to US publishers, but only scraped the surface of a growing industry.
And none are without controversy: Dennis Mortensen, founder and chief executive of New York-based Visual Revenue calls the editorial ‘decision support’ platform his company has developed “Susan”… and expects a healthy dialogue with her.
“She’ll tell you what stories to run, and help you to retain them while they are still attracting interest; which one’s in tune with your audience, and which isn’t,” he says.
He says ad hoc decisions by online editors can lead to dangerously self-fulfilling results, and the platform – recently complemented by a module to ‘test’ headlines – can avoid that, helping an editor to base decisions on fact.
Typically, the advice will be grounded on analysis of information from at least two sources – crawling for story data and importantly following readers. And it will support a decision on which story to give prominence – whether its own or that of the editor – with suggestions about which stories should surround it on the page.
The relatively-new concept is already being eagerly adopted by digital publishers. One client, two-year-old information and opinion site The Blaze (www.theblaze.com) uses the system to support its focus on “deeply engaged and high-value readers”.
Specific real-time recommendations from the VR system are like having a second front page editor, according to editor-in-chief Scott Baker, quoted in a company case history. He cites the example of a story about presidential hopeful Mitt Romney’s tax returns: “I knew it was an important story that would resonate but I had no idea to what extent,” he says. “I put it in top stories at 10.20am and it stayed there until 2.30pm as per VR’s recommendation. If I hadn’t seen how well the position and article were engaging the audience, I would have taken it down well before and lost out on the extra value.”
The bottomline is a claimed 59 per cent increase in article views driven from the ‘featured story’ position alone… and more from other homepage elements.
And there’s the implied corollary that if a computer can make such valid suggestions, why wouldn’t you want to let it get on with the job of implementing them?
In Chicago, a company called Narrative Science is working on the next part of the picture: Software that writes stories, and is doing so at the rate of hundreds an hour.
From data.
The results aren’t eloquent prose, but they’re functional… and they are helping to provide coverage which publishers find it increasingly difficult to justify and finance. It’s hard to get examples, but ‘Wired’ magazine’s Steven Levy cites Forbes among a largely private list of publisher clients.
One adoption in the US is via the very technology-savvy Little League Baseball (www.littleleague.org) which has a host of smartphone apps including one based on the GameChanger for iPhone, which provides tools for supporters to report match information, pitch by pitch. Instructions on how to set it up are even provided by a synthesised voice.
Narrative Science’s piece of the action is to turn that data into a – usually highly-positive – game summary with which parents, grandparents and other fans can follow the action on the web. And produce full match reports virtually before players are back in the changing room. The reports are positive, incidentally, because the system is programmed that way.
Figures quoted suggest the number of such reports is already close to half a million a year and may triple in 2012.
In the company’s two-year history, there have been numerous examples, and financial information is a natural for the technology. The format collects data – adding statistics as they become available – finding angles and placing it in researched contexts, then structuring sentences using pre-chosen vocabulary and templates. Tone can also be customised, and the company says it could cover the stockmarket “in the style of Mike Royko”.
Mike who? Wikipedia will tell you that Royko was a local hero; a Chicago journalist with (at least) two claims to fame: He won a Pulitzer Prize in 1972 and later – while working for the ‘Sun-Times’ at the time of its acquisition by News Corporation, famously commented that “no self-respecting fish would want to be wrapped in a Murdoch paper”.
But back to Narrative Science: One of a founding trio behind the company is chief technology officer Kristian Hammond, who sees a future in which news publishing is dominated by computer-generated stories.
His own story includes challenging journalism and computer science students at a local university where he and a colleague taught, and seeing them nut out a programme which took match data and turned it into prose… picking a picture to accompany it and writing the caption into the bargain.
The suggestion is not that algorithms such as Narrative Science’s will take over the jobs of journalists – few enough attend Little League, apparently, anyway – but that they will ‘write’ reports which would otherwise not be produced… either because there is no viable market, or because of the speed with which the data must be handled.
Not that there will be fewer stories written by humans, but that there will be more written by robots. And as Hammond reportedly predicted, the possibility of a robot winning a Pulitzer in as little as five years.
Comments