Norwegian News Agency is betting on automation for football coverage

The to-do list of a journalist covering a story is constantly growing. As well as writing the news report, it is likely he or she will also have to take photos, film a short video, and think about how this story might appear on social media.

With this in mind, Helen Vogt, head of innovation at Norwegian News Agency (NTB), has found a more preferable alternative for the future: robot journalism.

NTB has automated its football coverage, with its match reports written by a robot journalist and published within 30 seconds of the end of the game.

Humans no longer check the articles before they are published, trusting the robot to be accurate after extensive testing and work.

"Why go to all the trouble of making a robot, which is quite expensive?", she asked, speaking at the GEN Summit in Vienna yesterday (15 June). "We wanted it faster."

It takes the robot between five and ten seconds to write the report – the extra seconds are taken up by the publishing system.

Moreover, with a robot journalist on the job, the number of matches the agency is able to cover can grow, helping it broaden its audience in the process.

"We can offer reporting on those football matches, even if it's only friends and family who are interested," she explained.

Robot journalists are also more accurate, and do not make the same mistake twice.

Other newsrooms around the world have adopted automation to various extents. The Associated Press has automated its quarterly earnings reports, Yahoo has automated Fantasy Football league reports, and the Financial Times recently pitted one of its reporters against an algorithm to see who produced the better story – the human won.

But when discussing robot journalism, it's important to acknowledge there are many levels of automation, and two in particular that should be distinguished.

The first is when a journalists writes a template, and the algorithm then fills it in with the relevant data.

The second form of automation is when artificial intelligence and natural language programming comes into play.

"When you add some intelligence, that's when it becomes pretty fun. The basic level [the template] is going to be completely commonplace," Vogt predicted, adding that most number-based stories will be automated within the next couple of years.

The robot covering football for NTB takes its data from a database of live scores manned by people, as well as a database of historical context. Its stories tend to be between six and eight paragraphs long, and the training required in the newsroom is usually centred around preventing mistakes from happening.

The robot might not make the same mistake twice, but it could make mistakes that would be difficult to imagine in advance.

For example, it once named a player who scored an own goal as the winner of the game. This story came before the robot was allowed to publish without an editor checking, so it did not make it onto the website, but Vogt explained football is a low-risk subject for experimenting with automation.

"There's no big disaster that can happen with football," she said, adding that even if the own goal story would have been published, they would have recovered from it.

At NTB, the concept is that automation will free up journalists' time to focus on other projects. Vogt quoted Bloomberg editor-in-chief John Micklethwait, who recently sent a memo to staff explaining why the organisation was planning to set up a team to explore automation.

"Done properly, automated journalism has the potential to make all our jobs more interesting," he said.

Vogt also believes the audience will welcome articles written by robots, as algorithms enable "much broader coverage when it comes to local news, and also to personalise that news more".

"The audience will get a better level of service, and that is also good journalism."

Free daily newsletter

If you like our news and feature articles, you can sign up to receive our free daily (Mon-Fri) email newsletter (mobile friendly).