NTB and AI: Customising content for local audiences

An integration between Superdesk, our award-winning open-source newsroom content management system, and iMatrics, an AI-enabled auto-tagging tool, will soon provide journalists a powerful new way to distribute content to audiences with more precision.

Last year, NTB, the Norwegian news agency, published an average of 150 unique news articles every day covering a diverse range of topics – from weather in the capital to politics abroad. Ensuring these articles reach the right readers requires a meticulous application of metadata.

Until recently, metadata tagging was done manually, a thankless end-of-the-day task foisted on the bottom rungs of the journalistic ladder. This meant that metadata, despite its value to modern news production, was often an afterthought. Sourcefabric has partnered with NTB to take the mundane out of metadata processing. The project is expected to be completed by mid-February. 

Here’s how it will work: When a journalist finishes writing their story in Superdesk, they click a button that launches the iMatrics algorithm, which analyses the text and returns a list of suggested tags. Journalists will check these tags to remove false positives, add ones the AI missed, or create a new tag if needed. When content is published by Superdesk, it will be coded with these tags – categories like people, places, organisations, events, objects, and topics – to help NTB subscribers micro-target delivery to specific audiences.

Additionally, iMatrics-enabled content will be better aligned with SEO best practices, and will help the news agency – which produces content in the official written languages of Bokmål and Nynorsk – localise stories for a population of 5.4 million on a landmass 33% larger than the United Kingdom.

To assist in the iMatics integration, Sourcefabric is working to wire NTB’s Superdesk with an API so that the CMS and iMatrics can talk to each other. We are also helping to build a feedback loop, whereby published content will be returned to iMatrics to compare the AI recommendations with the final output. In this way, iMatrics’ software will constantly learn to become more accurate.

When launched, there will be myriad applications for this integration. For instance, automated business or weather stories could be tagged by geographic location and then delivered only to media outlets in the regions referenced in the content. 

National news stories could also be made more relevant. Let’s say the news agency was covering the annual Birkebeinerrennet cross-country ski race, which draws participants from around the country and beyond. With iMatrics, NTB stories and photos could be automatically tagged with skier names and hometowns, and editors could distribute this locally-relevant content to specific publications. The need for features like this will only increase as NTB grows its content output through AI story production.

Magnus Aabech, an editor in NTB’s technology and development department, explains the project this way: "To ensure customers can find the stories they are interested in, we need to have better metadata to help them filter during the distribution process. But to keep from overburdening our journalists, we also need to automate the process. The iMatics-Superdesk solution will do both."

The auto-tagging project is part of a larger metadata overhaul at NTB. Another piece in development is the consolidation of metadata terminology into a single "vocabulary," so that the tags used are consistent across desks and products – from press releases to sports data. Aabech says this will help clients map their own tagging structure to the NTB structure, which will further streamline content distribution and reception.

"We have about 170 media clients, most are local newspapers, and none are interested in all of the news we produce," says Tor Kristian Flage, Head of Projects at NTB. "Better metadata will help our clients select content that their readers want. That’s why having the right metadata is vital to our content distribution strategy."