Hyperlocal’s Automated Future

Share this:

I recently read with great interest a TechCrunch article about the clever combined offering of GameChanger Media and Narrative Science. GameChanger builds nifty statistics software that  for handheld and tablet devices that makes it easy for coaches or avid parents to capture all the action as data for future analysis. Focused primarily on baseball for now, this software makes it super-easy for coaches to collect detailed stats on player performance. It’s in heavy rotation among Little League, high school and college teams. Narrative Science is the startup that has built a journalism machine, a software product that can literally take raw data and craft original content.

Launched out of Northwestern University by two computer scientists who also are professors of journalism and a former senior executive of DoubleClick, Narrative Science has already scared ink-stained wretches while producing highly credible content and articles from box scores and other pieces of detailed data. I’ve read articles written by Narrative Science that have been virtually indistinguishable from simple, somewhat artless articles written by human scribes. The rub is that now a coach can push a button on his GameChanger console at the end of the game and Narrative Science will create a brief AP-style article recounting the results of the game. Here’s one.

It’s entirely possible that the only way hyperlocal — where the value is in the information, not the presentation — will truly succeed is to reduce the cost of content dramatically.

Now, let’s step back for a second. The biggest problem for hyperlocal media is creation of cheap content. Not great content, because, let’s face it – most of the articles written for micro-locals have never been the stuff of New Yorker Magazine or the Wall Street Journal. Just relevant, cheap content. For example, one of the most entertaining parts of my local paper is the police blotter. I learn a lot about what’s really going on around town. The police blotter is, of course, data. Narrative Science could easily turn that into an article.

Even better, Narrative Science can add context that would make it possible to craft whole  stories about each piece in a police blotter if, say, enough relevant information is accessible on the Internet or in accessible databases. So a blotter item about a bicycle theft becomes a story that includes a bit about a rising tide of bike thefts in town – constructed automatically. Government data is also quite available and detailed from the U.S. Census bureau down to zip code levels. That’s fodder for interesting stories. Voice-to-Text technology has gotten extremely good. At present, it’s only reliable for automated transcription of one speaker at a time. I am fairly confident that accurate transcriptions of simultaneous speakers may not be far away. At that point, City Council or School Board meetings become data and easy to turn into narratives.

Where it gets really interesting is the idea of Narrative Science working in tandem with something like Thomson OpenCalais, a powerful piece of contextual analysis software used regularly by ProPublica and the New York Times. Match the automated analysis of OpenCalais – which can, for example, discern the number of times a specific person is mentioned in a huge stack of documents and provide some context around the mention – with the storytelling of Narrative Science and you might get enough data to automate investigative stories or even crank out a whole series based on findings. Think, for example, of the possibilities of applying this to the Palin emails (I’m sure the NYT is thinking about this right now).

So what does this have to do with hyperlocal? It’s entirely possible that the only way hyperlocal will truly succeed is to reduce the cost of content dramatically. The way it’s been done to date is to entice people to write for free, to pay very low rates for posts ($50 for a post that might take a few hours to write and report), and to hire editors that are also paid low rates and given ridiculous work loads. Trust me, I lived it – it’s fun if you are young and hungry but gets old fast. If you can automate the story creation, then the editor can become a more highly paid curator and guide who sets direction for hyperlocal coverage, lets the machine do the writing, and then steps in to add high-level context (since they know where the bodies are buried) or style improvements.

At the hyperlocal level, the value is in the information, not the presentation. You read the local to learn, above all, what’s going on in your town or your nabe. If a computer can help collate and present that to you in a more digestible fashion, more better. Will this kill the community journalist? I doubt it. The journalist still must be present. Narrative Science can write up all the Little League stories but it will never have “story judgment” or a “nose for news.” Or an eye for an off-beat lead, a story hook that maybe is unconventional. Those are human traits and will still be valued, even if the journalist becomes more curator than creator. For me, I’d be ecstatic to get more and better local news information, even if it’s in slightly hackneyed prose. It’s news that impacts me directly and I want it from a computer, particularly if I’m not getting enough of it from the human scribes.

Alex Salkever’s Personal Fight column appears every Friday.