Home

The APQC Blog

That Machine is a Better Storyteller Than Me

I am willing to bet that machines wrote some of the emails, spider crawls, and media alerts that come across your screen every day. A grizzled sports writer didn’t write the recap of the baseball game; most weather alerts and forecasts weren’t written by a meteorologist; financial analyses and stock market reports were not written by an intern or junior staffer. A “smart” piece of software told a machine to write it.

While you weren’t looking, smart machines started doing “natural language generation” which we once thought only people could do with any credibility. Hooray, because many “knowledge-worker” writing tasks are grunt work inflicted on lower level staffers who can’t add a lot of value because they don’t yet have the business acumen. Or worse, our smartest experienced people are wasting their time generating boiler-plate for monthly reports that a machine could do instead.

As part of my ongoing quest to understand how ever-smarter machines are going to transform how we work and live, I interviewed Stuart Frankel, CEO of Narrative Science, one of the leading firms in the space of advanced “natural language generation.” (NLG)  (see photo above)

Narrative science as a field has been around since the Seventies when the Weather Service started using it to take today’s weather data and turn it into snippets of text for the weather reports. Now, using much more advanced software, Narrative Science clients include organizations like Credit Suisse, USAA, MasterCard, American Century Investments, and the U.S. Intelligence Community.

Stuart maintains that machines can now do a better – not just cheaper – job than humans of transforming data into stories. And we all know how people love a good story.  In fact, automated storytelling around the patterns in the data is the reason APQC’s Advanced Working Group was so interested in NLG in our ongoing research into cognitive computing. It could reduce the time employees spend trying to make sense of data or reporting it to executives.

Stuart: What we do at Narrative Science is “advanced” natural language generation. In essence, the technology and associated concepts have moved from taking data and turning that into little snippets of text to a point where it is now a technology that can actually look at data, figure out what’s interesting and important in that data, and then take that analysis and render that into a natural language document. For example, a natural language document can be a 20-page investment research report that looks and sounds like a Wall Street equity analyst took a look at a stock, conducted research, and wrote up a report—a significant advance from where natural language generation was in the early days.  Ask yourself, “What is that data telling me that I can actually act upon and make decisions against?” I think that’s the heart of data storytelling.”

Stories Make Data Come Alive

“Every data-set, every database, every spreadsheet has a story to tell…”   - narrativescience.com

I have always said that stories make data come alive. If you only have one minute to get someone’s interest, give them the one-two punch: “Did you know that X percent (data) of our employees visit the community at least once a week? Jane James loves to tell people that’s where she got the engineering specs that won us that last big deal with XYZ Corporation.”

From an executive perspective, I would enjoy a computer-generated report every Monday of how often APQC members visited our website (hint: www.apqc.org) and how that compares to last month or last year. I don’t want to ask our already busy staff to do that. Better yet, send the text report to Siri and let her read it to me on the way to work. Tada – that’s where natural language generation comes in.

The lesson is to ask ourselves “What repetitive data analysis and reporting could be off-loaded to a computer with the right software?”  What about:

  • Content consumption reports by various communities of practice or functions in or organization?
  • Patterns of mistakes or lessons learned reported last week?
  • Weekly, monthly and quarterly sales reports?
  • And so on.

How else are we going to tackle all that big data and make it fun to consume?

For more on my interview with Stuart, including what jobs are at risk as this technology matures, view the full transcript.

Check out the rest of my Big Thinkers, Big Ideas interviews on APQC’s Knowledge Base.

Subscribe to the Big Thinkers, Big Ideas podcast on itunes or on APQCPodcasts on Podbean.

You can connect with me on Twitter @odell_carla