How do you make a machine write like a human? Narrative Science, an artificial-intelligence company that can computer-produce analytical reports, has one answer. Imagine an organization has just hired a new writer. Her boss discusses the topic of coverage with her and passes along a style guide. She writes a bunch of samples. She gets feedback. She writes some more. At some point the feedback stops because she knows what she’s doing. According to Kris Hammond, Narrative Science’s technology head, that’s exactly how his company’s technology, called Quill, works. “It just happens to be the case that it’s a machine,” he says.
Quill achieves this through a series of logical steps that in a human would be called clear thinking. “Who is going to be reading this story? Why? What do we want to communicate? That gives us a snapshot of what information you need in order to tell the story,” Hammond told Quartz. It’s a method that any writer—human or machine—would do well to follow.
Big data = too much data
Last month, Narrative Science raised $11.5 million in funding, taking its total up to some $21 million. Calculating a precise figure isn’t possible since one of the previous investors is notoriously tight-lipped: the US Central Investigation Agency. That got Narrative Science plenty of attention. In fact, two-thirds of Narrative Science’s business comes from non-media companies. While a handful of publications—most notably Forbes—use its software to write quick pieces, its technology is better suited to internal reports.
The CIA, for instance, produces vast quantities of intelligence that its human analysts would take eons to analyze. Large companies can use existing technology to gather sales data but they need people to sift through all of it. Indeed, in an age of “big data,” even medium-size companies are producing so much information that understanding it can be a challenge. With existing technologies, “you can export a spreadsheet or see some charts or graphs. But we can write the performance review. For each individual sales person, for their management, for upper management,” says Hammond.
It’s a seductive idea. As is the notion, espoused by Narrative Science, that infographics and fancy charts pose no threat to a succinct, well-written piece of language. Hammond gives the example of an analyst who digests information from charts and spreadsheets and translates it to natural language for his bosses. That’s the job done by Quill.
The next step
Narrative science is three years old with high-profile clients and has plenty of cash in its war chest. With the latest round of funding, it’s looking to enter a “rapid-growth phase.” Hammond says the company has proven its technology works. Now it wants to develop into something that is seen as essential to businesses. “Google’s job is to index the world,” Hammond says. “Narrative Science’s job is to make sense it.”