Narrative Science Employs Natural Language Generation
If you’ve spent much time on Nanalyze, you know that we’re passionate about technology and believe that we’re living in the most exciting times in history. We’re talking cure-for-cancer and sending-humans-to-Mars territory here. Our job is to keep you up-to-date about these changes in a variety of fields, so you can make informed financial decisions about where to invest or not—and learn some pretty cool stuff along the way. We talk about the good, the bad and ugly no matter what.
Then we came across Narrative Science and its natural language generation (NLG) platform Quill, which uses artificial intelligence technology to write everything from financial reports to sports news. In other words, it’s the competition. That’s right: Looks like we’re going from gonzo journalism to robo journalism. We knew that English degree would be obsolete someday.
Narrative Science Employs Natural Language Generation
Chicago-based Narrative Science came onto our radar this month when the gurus at CB Insights selected it for the Artificial Intelligence 100 list at the Innovation Summit. However, the company has been around since 2010. Since then, the AI startup has raised nearly $30 million, the most recent infusion of cash coming from a $10 million Series D in November 2014. Among its investors are Battery Ventures, Sapphires Ventures and USAA, which is also a customer.
Many of its 70-plus enterprise customers, in fact, are in financial services, such as MasterCard and Credit Suisse. Another 12,000 or so small businesses use Narrative Science’s self-service applications. Media companies are also among their customers, using Quill to produce data-driven articles. Take a look at the screenshot below from Forbes. Note the author bio (though it appears that Narrative Science hasn’t produced a story in this space for more than a year).
Natural language generation
You may remember that not so long ago we waxed philosophical about aspects of so-called robo journalism in a pieced we called Content Creation by AI Powered Robot Writers. While we conceded that computers are creating online content, we thought the AI bit was still some ways off.
Narrative Science insists its “advanced” natural language generation platform is a subfield of artificial intelligence, producing language as an output of data. It’s sort of the opposite of another AI topic that we often cover—natural language processing. In simple terms, NLP analyzes language to extract data, while NLG does the opposite by analyzing data and turning that into language.
Other companies, like Automated Insights (more on them below), also use natural language generation but rely on templates to generate stories. Narrative Science says Quill can be trained to mimic the style and tone of its customers, spitting out texts in all sorts of formats, from bulleted lists to long-form narratives to tweets and web content.
As Narrative Science CEO Stuart Frankel wrote in a little publication called the Harvard Business Review:
“Advanced NLG platforms … start by understanding what the user wants to communicate. Then these systems perform the relevant analysis to highlight what is most interesting and important, identify and access the data necessary to tell the story, and finally deliver the analysis in a personalized, easy-to-consume way: as a narrative.”
Formerly a research project at Northwestern University called StatsMonkey (a software prototype that automatically generated baseball game recaps), Narrative Science today claims 11 patents. It has a number of partnerships with companies that specialize in analytics and visualizations, which integrate Quill into their platforms. For example, Microsoft Power BI, a suite of analytics tools, uses Narratives for Power BI to generate narratives that explain insights from data and visualizations. Software company Tableau does something similar with its suite of business analytics products. In this case, Narratives for Tableau is a Google Chrome extension that automatically generates stories about visualizations created in Tableau. See the video below for more about how this works:
Research technology firm Gartner predicts in just two years that natural language generation will be a standard feature of 90 percent of modern business intelligence and analytics platforms. In a recent article, Gartner says, “Currently, BI teams integrate stand-alone NLG engines, but as the technology evolves, that will change. NLG will enable next-generation BI and analytics platforms to automatically find, visualize and narrate important findings. The technology will expand analytics to a broad audience as well as reduce time and cost for regular batch reports.”
As you’d expect, Narrative Science isn’t alone in developing natural language generation platforms to produce weather reports, news stories about your fantasy baseball team or quarterly earnings reports. We’ve found companies in Germany, Spain and even Russia, but we’ll be American- and Anglo-centric for this article.
We briefly mentioned one chief rival, Automated Insights. A startup founded in Durham, North Carolina in 2007, Automated Insights was snatched up by private equity firm Vista Equity Partners in 2015 for $80 million, according to Business Insider. The deal came on the heels of another acquisition by Vista Equity of another NLG competitor, STATS, which focuses on sports content. “Vista’s resources and STATS’ distribution will allow us to fast-track our Wordsmith natural language generation (NLG) platform across multiple industries,” Allen wrote on the company’s blog at the time of the buy-out.
Wordsmith is Automated Insights’ equivalent of Quill by Narrative Science, though the former uses templates populated by the client to spit out paragraphs, meaning less AI in your AI robo writer. It hits many of the same industries—finance, sports and media—as Narrative Science. And it also sports some big-name clients like the Associated Press (traitors!), Cisco and Yahoo!
Automated Insights states that they are “the only NLG company that has a platform that customers can actually use on their own”. When they created Wordsmith, they re-imagined NLG as something that could be used by the masses and implemented easily. In some cases, their customers have implemented the solution is just a few days time. They’re now growing at a faster rate than before having tripled their customer base and generated more than 1.5 billion pieces of content last year, a number that they say is a couple orders of magnitude more than their competitors.
A publicly traded British company on the London Stock Exchange, Arria NLG (LON:NLG), is more of an apple-to-apple comparison to Narrative Science in terms of AI natural language generations platforms. Like Narrative Science, Arria began life in academia. On Dec. 13, 2013, however, it opted to accelerate the startup part of development—it was founded in 2011—and went IPO. It had a spectacular opening at about $140 per share. Today, it’s trading under $8, near its 52-week low of $6.81. We wonder what algorithm had to write that story.
Apparently the stock crash was partly due to a contract break with Arria’s main client, Royal Dutch Shell, in mid-May 2015 that was worth up to $10 million. The company had managed to lose almost all of its £100 million valuation between December 2013 and May 2015, according to the website Interactive Investors. However, it’s managed to stay afloat, convincing investors over the last year-and-half that its product, Arria NLG Platform, is solid.
Arria NLG recently launched a new product, Recount, a SaaS-based solution for small- and medium-sized businesses. Arria says Recount provides these companies with “access to full-time skills common to financial analysts, spotting trends, identifying problems, and forecasting what’s likely to happen next.” And, of course, it tells stories using natural language generation, along with interactive charts, detailed reports on sales, cash position, expenses and simple sales forecasts. Sounds like a good story—but we’re still not buying the stock.
Is it becoming clear that natural language generation platforms are better at generating headlines about the demise of human journalism than actually producing journalism? We’ll concede that they can write 60 daily weather reports as well as a meteorologist, not to mention 120,000 times faster. But can they ask hard-hitting questions? Infuse narrative with subtle humor and warmth? Drink like a fish and bang out a thousand words with a tequila hangover? We didn’t blast Hunter S. Thompson into outer space because he could spell well.
At this time, the real value in natural language generation platforms, as we see it, seems to be in providing narratives to help understand financial data, visualizations and the like. The obvious trend, as Gartner notes, is the integration of NLG into analytics and visualization products. The acquisition of Automated Insights by Vista Equity could herald a broader theme of folding NLG startups into bigger portfolios: another AI cog in the machine. But we’re not quite ready to write off companies as nimble and promising as Narrative Science.