It’s been some time since I last had a look around at where people are at with robot journalism. The companies that still tend to come to mind (for me) in this area are still Automated Insights (US, Wordsmith), Narrative Science (US, Quill), Arria NLG (UK), AX Semantics (DE), Yseop (FR, Compose) and Tencent (CN, Dreamwriter) (was it really four years ago when I first started paying attention to this?!) but things have moved on since then, so I probably need to do another round up…
A recent addition I hadn’t noticed comes from the Washington Post, (owner: a certain Mr Jeff Bezos), and their Heliograf tool (via Wired: What News-Writing Bots Mean for the Future of Journalism); this currently seems to write under the byline powered by Heliograf and had its first major outing covering the 2016 Olympics (The Washington Post experiments with automated storytelling to help power 2016 Rio Olympics coverage); since then, it’s also moved into US election reporting (The Washington Post to use artificial intelligence to cover nearly 500 races on Election Day).
The model appears, in part, to be using Heliograf as a drafting tool, which is one of more interesting ways I always thought this sort of stuff would play out: “using Heliograf’s editing tool, Post editors can add reporting, analysis and color to stories alongside the bot-written text. Editors can also overwrite the bot text if needed” (my emphasis, via).
It seems that automated election reporting has also been used in Finland. Via @markkuhanninen, “[o]ur national broadcasting company Yle did the same for elections and some sports news (NHL ice hockey). All articles seem to be in Finnish.”
(Hmmm… thinks… maybe I should have tried doing something for UK General Election? Anyone know if anyone was using robot journalists to produce ward or constituency level reports for UK general election or local elections?)
Looking back to 2015, Swedish publisher Mittmedia (interview) look like they started off (as many do) with a simple weather reporting back in 2015, as well as using NLG to report internally on media stats. The Swedish national wire service TT Nyhetsbyrån also look to have started building their own robot reporter (TT building “reporter robot”) which perhaps made it’s debut last November? “The first robot reporter has left the factory on @ttnyhetsbyran and given out in reality. Hope it behaves” (translated by a bot…) (@matsrorbecker).
And in Norway, it looks like the Norwegian news agency NTB also started out with some automated sports reports last year, I think using orbit.ai? (Norwegian News Agency is betting on automation for football coverage, and an example of the recipe: Building a Robot Journalist).
2016 also saw Bloomberg start to look at making more use of automation: Bloomberg EIC: Automation is ‘crucial to the future of journalism’.
Offhand, I haven’t found a specific mention of Thomson Reuters using automation for producing business reports (although I suspect they do), but I did notice that they appear to have been in the document automation game for years with their Contract Express application, a template solution that supports the automated creation of legal documents. A quick skim around suggests that document automation is a commodity level service in the legal industry, with lots of players offering a range of template based products.
Thinking in terms of complexity, I wonder if it’s useful imagining automated journalism in the context of something like: mail merge, document automation (contracts), automated reporting (weather, sports, financial, election, …)? Certainly, weather reporting and sports reporting appear to be common starting points, perhaps because they are “low risk” as folk get comfortable with producing and publishing automated copy.
I also wonder (again?) about how bylines are used, and have been evolving, to attribute the automated creation of news content. Is anyone maintaining a record or collection of such things, I wonder?