Do you think a software program could perform your job?
Before you scoff, a U.S.-based company has created an algorithm that takes data, like sports statistics, company financial reports, and housing starts, and turns them into newspaper articles.
The code is the work of Narrative Science, offering proof of the progress of artificial intelligence—the ability of computers to mimic human reasoning.
The New York Times wrote about the company this month: “For years, programmers have experimented with software that wrote such articles, typically for sports events, but these efforts had a formulaic, fill-in-the-blank style. They read as if a machine wrote them…[but] articles produced by Narrative Science are different.”
Here’s an example written by the software:
“WISCONSIN appears to be in the driver’s seat en route to a win, as it leads 51–10 after the third quarter. Wisconsin added to its lead when Russell Wilson found Jacob Pedersen for an eight-yard touchdown to make the score 44–3 …”
Not too shabby, considering it was written by software.
The company apparently has 20 clients so far. On its web site, the company notes that the technology is being used for sports stories, financial reports, real estate analyses, sales and operations reports, and market research content. No mention of technical writing but I don’t see why it couldn’t be used for some documents.
I believe that the technology could be used for documenting bug fixes and new features that might appear in Release Notes. Of course, there are many things that a writer does every day—such as project management and interviewing—that would be difficult for a piece of software to emulate.
This article reminds me of a blog entry I wrote last year: “In short, some of the highly analytical jobs are becoming commodities that can be performed by a computer or an inexpensive worker in Asia.” See my earlier blog entry.
What do you think? Could a piece of software potentially write some of your content?
I have worked as a
Software has been writing technical content for years. I’ve created multiple systems that extract information from source code to build references. There are multiple systems on the market that automatically generate API documentation from a combination of the code itself and inline comments. There are tools that generate documentation for XML schemas.
This is not unique to technical writing. Contracts are often assembled from boilerplate based on data. Much of the content on an Amazon page is generated from data. Sports stories have always been formulaic. The only thing surprising about this story is that no one has automated them sooner.
There is absolutely nothing new about this, but a great deal more could be done, especially in technical communication.
You may have been using technology to create API documents for years. We’ve all used technology to assist us with our work.
But, as far as I know, there hasn’t been technology to actually write paragraphs of content that might appear in a user guide or online help system. I believe that’s quite different from your example
It is different. It is also different from the example in the article, which is just making sentences out of statistics. That kind of thing has been done for years. Their claim that they have found a way to make machine generated content sound more human is mildly interesting, but not really all that relevant in the kind of statistical reporting that they are applying it to.
When i use hammer and chisel, the better result that i can obtain is to broke the blocked lock of the door of my garage… With hammer and chisel, Michelangelo sculpted “Moses” and “The Pieta”, two masterpieces of the human history. SAME TOOLS… BUT DIFFERENT “HUMAN” BRAINS AND SOULS… I suppose!:-) We use many useful tools in writing activities… but i don’t think that a tool (ANY tool) can substitute the knowledge of a Technical Communicator. My customers ask a document clear, concise, appropriate, accurate, essential. Ask me videotutorials, ask me graphic contents. Ask me to explain easily very complex concepts. I’m sitting on the bank of the river… when will arrive a tool able to compose automatically all this.. call me!:-)
Thanks for your comment Alessandro! Interesting feedback.
I tend to agree that a fairly linear list like the problems fixed in this release and the What’s new section can perhaps be automated.
What I also agree with is that a lot of what technical writers do when they do not write cannot be automated. In particular, the thinking process of deciding what not to write because the product they created with their teams is so simple that it does not require any documentation at all 🙂
Interesting article, thanks for sharing.