One of the largest by-products of the digital revolution is data, and entrepreneurs are finding new ways to harness and make use of the increasing variety of data. In the following interview, Kristian Hammond, CTO of Narrative Science, talks about how his company generates narrative stories from gathered data — a function that could play out very well in content organizations such as newspapers, allowing them to scale content without having to hire more staff.
Hammond says stories grounded in data work best — think sports stories, to start — and that the increasing amounts and kinds of data being produced create new opportunities for the kinds of stories that can be generated — think pharmaceutical testing reports. He will expand on the ideas and concepts behind using data to generate content in the Scaling Content Development Through Automation session at the upcoming Tools of Change for Publishing conference.
What does Narrative Science do and how are you applying the technology to journalism?
Kristian Hammond: Narrative Science is a Chicago-based company that is focused on the automatic generation of stories from data. Spun out of the schools of Engineering and Journalism at Northwestern University, we are currently working with customers (in both media and business), generating content from public and proprietary data sources.
We are generating stories in the arenas of sports, finance, real estate, and politics. We are also working with companies to transform business data into client reports, franchise statements, and customer communication. In effect, we are giving a voice to the insights that can be found in the growing world of big data.
Our aim is to provide content and insight in those areas where it is either financially or logistically impossible for organizations to generate it themselves using traditional methods.
How does data affect the structure of a story?
Kristian Hammond: Our stories are driven by data, but they are not simple recitations of that data. In doing an earning story, for example, having (or not having) historical data will change the scope of the story. In the former case, we will have year-to-year comparisons; in the latter we won’t. In sports, seasonal data will allow us to give a voice to trends and rankings, and rivalry data will allow us to describe a game in terms of the impact of the game beyond the score and stats. In all of these cases, the greater the pool of data we have, the more powerful the story will be.
What kinds of stories lend themselves well to this type of system and why?
Kristian Hammond: The technology is designed around transforming data into stories. Stories that are themselves grounded in data are the perfect match for us. As more and more of the data that defines our world comes online, we see more and more opportunities to create new stories in new domains.
What kinds of stories just won’t work — what are the boundaries or limitations?
Kristian Hammond: Often, stories are the products of long-term observations, conversations and ongoing inquiries. A story in Vanity Fair that is the product of 30 conversations, for instance, is not something we would ever try to do. Also, stories that are more opinion based are outside our realm. But again, as more information is transformed into machine-readable data, there are more opportunities for us to use that data to expand our realm of possibility.
In what ways can publishers benefit from Narrative Science?
Kristian Hammond: Publishers who are resource bound or who want to expand the scope of their reporting are perfect clients for us. If a financial publisher is producing earning previews for 30 companies, for example, they can use this technology to generate exactly the same kind of story with the same tone and language for 1,000 companies. If a publisher wants to track real-time events and there is data around them, they can use us to generate everything from stock alerts to in-game quarterly summaries. Wherever there are problems of scope in terms of volume or the constraints of time, publishers can use us to create the stories they simply do not have the resources to write.
In what other industries are you finding applications for Narrative Science?
Kristian Hammond: While Narrative Science began its life providing content for media companies, it has expanded its reach to cover reporting for all types of organizations that have data describing their businesses and operations. We currently provide reporting for client services, tracking franchise operations, and performance reviews for a variety of companies. We are also looking at how our platform could be used to transform the huge data repositories captured in pharmaceutical clinical trials into clear and concise reports that provide overviews of and insights into their results.
In effect, anywhere there is data and a story to be told from it, our analytics and narrative generation platform can leverage that data into insight.
This interview was edited and condensed.