To free up its reporters to produce more in-depth coverage of the Olympic Games in Rio, the Washington Post has been automating updates about results and the event schedule.
Using a system built in-house called Heliograf, around 330 articles and 600 alerts about the Olympics are automated, published both internally for the benefit of reporters, and publicly to keep readers informed of the latest results.
"In past Olympics we have had sports staff members, reporters, editors and producers, who watched the live feed, watched TV, to write down the results of various medal events," said Jeremy Gilbert, director of strategic initiatives, the Washington Post.
"That took a fair amount of time. We wanted that as a service to our readers, to be able to tell them right away who won if they were interested, but we also know that's pretty publicly available data.
"We would much rather have our reporters, editors and producers working on stories that only the Washington Post can have instead of something like that, that anyone can witness."
When setting up the Heliograf templates that power the Olympics coverage, the Washington Post aimed to automate three elements: publishing the results of each event, alerting readers to upcoming events, and providing medal tallies for participating nations.
Once the IOC provides the results data for each event, it takes the system around 45 seconds to create the post, which is then published to social media including Twitter and the Facebook Messenger bot, and to the Washington Post's running live blog.
"The time to publish is almost exclusively the time it takes the IOC to move the data," explained Gilbert.
For some events, such as swimming, this process is almost instant, whereas for others, it can take quite some time.
In certain cases, like the cycling road race, the data is only made available after every athlete has completed the race, which could be over 30 minutes after the medallists have crossed the finish line. This can sometimes put Heliograf at a disadvantage.
"Our side of the process is relatively quick and relatively straightforward," he said. "So the work in automating stories for the Post is in the creation of templates and the logic that powers those templates much more than it is in the act of importing the information and putting out the information."
Even as we make these tweaks to the Olympics system, the development team is very focused on what we can do for November.Jeremy Gilbert, The Washington Post
Once the automated updates are published, they are checked by human reporters. The Washington Post never envisioned an automation process that would require a human's approval before publication, so the system is built with certain error-checking elements in place.
Heliograf would recognise for example if the results were unlikely – say, if they showed an athlete established a new world record by a great difference, which could signal an error in the data.
"What we are trying to do as much as possible with this system is capture the logic that our editors are using when they make assessments."
The Washington Post's sports reporters were heavily involved with the creation of Heliograf's templates for Rio coverage.
"Hopefully what we're doing is then freeing them up to cover how someone could win, why Katie Ledecky's swimming style is so powerful, any of the things that can't be determined just by watching the event on the screen or just by looking at the data."
Heliograf was tested out of the public eye during the primaries and caucuses, and the Olympic Games are an opportunity for the Washington Post to improve the system before the US presidential election on 8 November.
Four reporters worked overnight in 2012 to deliver "rogue versions" of results during the last election, so the automation efforts at the Post were prompted by a drive to improve editorial workflows on election night.
"Wouldn't it be great if we could not only tell those stories but we could use some mathematical analysis, some historical analysis of those election results, and publish them instantly instead of having to work all night and publish them next morning? That's where we really started," explained Gilbert.
Incremental improvements are being made to the system over the course of the Olympic Games, and the goal is to further develop it and integrate it into the Washington Post's content management system, Arc.
"Believe it or not, even as we make these tweaks to the Olympics system, the development team is very focused on what we can do for November. We only get one shot at it, and we need to make sure the system is perfect."
Free daily newsletter
- Martin Baron: 'There is a business case in addition to the moral case for investigative journalism'
- Advice for young journalists from Martin Baron, executive editor, The Washington Post
- The Washington Post has a Facebook group where readers can ask reporters how and why they cover stories
- The Washington Post is using augmented reality to let audiences explore iconic buildings with their iPhone
- 3 automated transcription tools for journalists