Month: June 2015

At the halfway mark in our eight-week Structured Stories project on Friday, Natalie, Ishan and I decided to measure our performance.

By the numbers, we’ve made substantial progress — we’ve created 182 new events in 15 stories, all of which are now live on the Structured Stories website.

The more events and stories that we input, the more we find that our thinking about narrative stories changes. Increasingly, we notice ourselves deconstructing the news as we read it, breaking down articles into a series of finite events, and dicing those events into their primary nouns and verbs.

We’ve learned not to worry about engaging leads or colorful language. Instead, we focus on crafting clear, concise and specific events that are easily “structurable,” to use a term recently coined by David.

We are, in other words, finally beginning to think like structured journalists.

But a number of questions remain. In fact, sometimes it feels like the more progress we make, the more questions — big and small, technical and editorial — we have.

We’re helping David make improvements to the content management system. As we input events for our four main topics — policing, bail reform, housing and Uber — we’ve found more than 25 bugs. The list of unresolved editorial issues currently stands at 56 — a number indicative of how much we’ve learned, but daunting nonetheless.

One of our most persistent struggles remains translating events we intuitively understand in language to structured events.

In a traditional article, for example, it makes sense to say that airports have started ticketing Uber drivers. In a structured story, however, this statement would have to be attached to a specific event — with a specific authority, time and place.

We’ve tackled issues like these in hours of daily check-in Skype sessions with David, countless messages to David on Slack and near-constant discussion among ourselves.

David has patiently reassured us that this question-filled dialogue is not only natural, but also helpful in the long term. He’s reminded us that we’ve used language for tens of thousands of years, but that this data-driven approach to narrative is still nascent.

“Finding an alternative to using language in writing is a pretty audacious goal,” he noted. “It makes sense if it feels a little weird, a little unnatural at first.”

Comments closed

We hit the ground running last week, eager to begin constructing structured stories on our topics.

I Googled everything I could about “NYC Housing” and was quickly overwhelmed. For some context, I stopped by City Hall for a hearing on the New York City Housing Authority’s plan to erase its deficit. Rachel researched Mayor de Blasio’s relationship with the police, and pieced together the myriad of events that soured their bond. Natalie tracked Uber’s meteoric rise and the subsequent PR nightmares that engulfed the ride-sharing company.

We hit our stride after a few days and marched through a routine: research, input an event in a structured story, repeat. It was slow work, especially if we had to verify conflicting accounts by checking primary sources.

“There’s something noble about making sure everything is correct,” said Natalie, alluding to our satisfaction once we solved something and could (finally) move on. Every day, we managed to finish dozens of “events,” the individual units that form the backbone of Structured Stories.

But, in a testament to how different this work is from traditional journalism, we are still having difficulty adapting to the unique writing requirements.

A structured story is different than “regular” writing because it’s all about breaking the news into data.

The data comes in two flavors: verbs and nouns. Verbs can be linked back to the FrameNet database, an expansive project that tracks meaning. Amazingly, the FrameNet database can be read by both humans and computers. It translates complex human meaning into data.

Nouns come from Freebase, a sprawling database owned by Google. Freebase assigns items unique identifiers, and we use these IDs to track characters or topics over time. De Blasio, for instance, is known in the database as  /m/0gjsd3.

In Structured Stories, combining verbs and nouns creates a data-rich event. And that data can be manipulated, allowing readers to see links between stories or track events over time. That’s the power of structure.

David Caswell, the creator of Structured Stories, told us our confusion was natural. A structured story in its raw verb/noun form is not meant to be read by a human. In fact, most readers won’t see the structured view when they visit the Structured Stories platform. They’ll read the bullet-points or summaries, which Rachel, Natalie and I write after we have structured an event. Bullet-points and summaries are the “normal” human sentence behind an event. Underlying that sentence, though, is a web of connections and malleable data that will provide readers with new information they have never been able to get before.

This project makes me feel like I’m learning to write again. I’m paying extra attention to nouns and verbs and stripping events to their core meaning. There seems to be a constant tug of war between language and structure when writing these events, with the ideal falling somewhere in the middle.

For now, we’re still searching for that happy medium.

Comments closed