Live from the Supercup: DFL uses AI to scale content creation
AI-based processes are being increasingly applied to expand content across the Deutsche Fußball Liga’s (DFL) app, website and social media channels. In particular it is being applied to scale and to personalise content published as Bundesliga Shorts, a sequence of images or videos in smartphone-optimised vertical format. Every image is overlaid with a brief, descriptive text, and users browse through the story image after image.
Since the story format requires much more visual material than conventional articles, along with a different layout, it comes with an increased need for editorial resource. “This has prompted us to search for an AI-based approach to supporting the production of stories,” says Björn Rosenthal, head of product at DFL Digital Sports.
He has led an effort to supercharge editorial with generative AI, continuing a tech partnership with AWS, and the results will be rolled out across the DFL’s app in less than a month.
To generate text, the DFL’s content management system uses the large language model ‘Claude’ offered by AWS partner Anthrophic. “All this runs on AWS’ AI service AWS Bedrock, which we are using,” says Rosenthal. This service is now embedded in the content management system.
Data collected from DFL matches – amassing over 3.5 million data points per game – is already being utilised by DFL subsidiary Sportec Solutions to feed the Match Facts statistical and analysis feed. The same metadata is also being fed into an algorithm along with the timestamp of the TV footage to extract individual frames showing specific match events for stories content production.
The system can automatically search for images using computer vision and AI and generate text components. Further algorithms can then sharpen the pictures, but Rosenthal emphasises that human editors have the last word to check or refine before publishing.
“The source written content is always from an editor,” he says. “AI allows us to piece together all the assets for each story faster, and from that one original source story we can generate versions in multiple different languages and target stories according to target demographics.”
As an example, he cites targeting a 22-year-old male soccer enthusiast in Wisconsin with a different version to what would be targeted at a Japanese female. “Right now we have a German and an English editorial team,” he explains. “To be more present in other markets, we will offer our content in a range of additional languages – such as Brazilian Portuguese, Spanish and Japanese.”
AI is also being used to enhance media production in ways that remain in test phase.
Another advancement, working with AWS, has been improving the discoverability of content in the DFL Media Hub, considered the world’s largest archive of football footage. “We see major potential in being able to broadcast content in several languages within a narrow time frame and are using AI to accomplish this,” Rosenthal says.
AI-generated metadata allows users to search for content in more than 210,000 hours of video footage more efficiently. Rosenthal suggests that host broadcaster Sportcast and domestic and international broadcaster partners would be allowed access to this content to enhance their live coverage of Bundesliga matches.
“You could type prompts into the AI for the type of information or clip you need and the AI would search the archive and serve it up to you instantly for insert into the live feed,” he suggests.
“We see major potential in being able to broadcast content in several languages within a narrow time frame and are using AI to accomplish this”
The DFL says it owns the data on which it is training its AI models. Steffen Merkel, DFL CEO, says he didn’t anticipate any tension between the league’s exclusive ownership of performance data and the players themselves.
“We collect and own match data and, to me, it’s a key prerequisite to making the media product better today. That players might challenge this in future does not concern me at the moment.”
Simon Rolfes, sports director at Bundesliga team Bayer 04 also keeps his eye on AI for possible use in crunching data for scouting and match analysis
“I am very interested in tech but I’m also critical. I believe tech will only survive if we use it. I’m open to testing everything, but if we do not use it, we will kill it. If it’s not useful we will stop the project.”