Technology is already being used by companies like the Associated Press and Yahoo to auto-generate data-heavy articles about quarterly earnings, college sports, and even fantasy football recaps. Today, the company is unveiling a public-facing version of its Wordsmith platform for anyone to use. You can sign up for access to the beta, with general availability being planned for sometime in January.
The Wordsmith platform is designed to automatically generate natural language reports based on large data sets. Within each project (defined by the data set, uploaded as a CSV file), you can create multiple “narratives,” which are in a sense high-powered Mad Libs. Write the basic structure, swapping key words for the variables available, and then add logic.
The key feature of Wordsmith is how it enables you to create branching paths, conditionally adding / modifying certain words, phrases, or entire sections based on the data available. Is a 60-inch TV considered “large,” “massive,” or “run-of-the-mill”? If Company A misses market expectations by $1.2 million, is that a massive failure, a minor hiccup, or something else entirely? Branches can also be embedded within other branches, allowing for more complexity, and you can choose to “randomize” what phrase is presented at any given time, for sake of variety.
It may seem like a lot of work for one, two, or even a dozen articles, but Wordsmith is intended for much larger scale. Once you have a template, all you have to do is update the data to generate new articles. (The AP, for example, publishes more than 3,000 financial reports each quarter using its version of Wordsmith.) The work, then, becomes less about writing individual reports as it is maintaining certain templates and narratives, updating for sake of creative variety and editorial standards.
“The writing process has largely been untouched,” Automated Insights’ CEO Robbie Allen tells us over the phone. “The way that you write now is not much different than how you did it 30 years ago.” It’s isn’t as much about commoditizing writing as it is admitting that so much of what we write is rote and repetitive (as the AP told us in January, no jobs have been lost as a result of its “robot journalists”). Allen said he’s interested to see what non-conventional uses people find for it, citing a Game of Thrones battle report generator that some employees made during an internal hack-athon.