Swedish Radio has deployed automatic recommendation technology in its newsroom to help journalist leverage the value of the organization’s vast and ever-growing content library. SR chose Peach, the EBU Member-developed suite of personalization tools, as the engine for their internal recommendation algorithm, building on an existing integration in the company’s digital on-demand services. We spoke with SR’s Head of Digital News Development Olle Zachrison and Personalization Product Owner Jörgen Bang about their experiences and plans for the future.
EBU Technology & Innovation: First things first – how do you use Peach in the newsroom?
Olle Zachrison, Head of Digital News Development at Swedish Radio: For a modern and digital newsroom, having access to an effective and continously developed recommendation engine is key. At Swedish Radio we have an audio-focused strategy for our digital platforms, which historically has made recommendations harder as a lot of our most important content only exists as sound. But by combining speech-to-text technologies with Peach we found a way forward. Today we use Peach to run through our transcripts and serve our digital editors with much better results than they would be able to get without it. This saves time, for us in the newsroom, as well as for our audience, which is increasingly impatient and expects us, as public service media, to deliver relevant information and context.
Jörgen Bang, Product Owner for Personalisation at Swedish Radio: To illustrate the scale, we have 30 editorial staff members publishing an average of 350 news pieces per day, so that amounts to a lot of content over time. Peach has proven to be a very good way of making better use of this content. We were able to deploy Peach as a running system, ready for us to ingest data and receive value. This value is made available to our own journalists in the form of an assistive system that automatically surfaces relevant related stories on any given topic that a journalist is working on, directly in our content management system (CMS).
EBU T&I: Why did you choose Peach for this use-case?
Jörgen Bang: It was a bit like the use-case chose us! Like I said, the volume of stories we produce is very large. It was too complex for our tagging and classification system, which made it very difficult for our journalists to use consistently.
At one of the Peach sprint demos, we saw that another public service broadcaster was exploring news recommendations with interesting results. Around the same time we had just started to automatically transcribe all our news audio to text. So, Peach had the capabilities, and we had the data. First, we did a short exploration to see how Peach could produce recommendations of related stories that would give a news story more depth and context, and we saw that it worked. The Peach infrastructure is already in place, so when we decided to implement this fully – from news transcript ingestion to recommendation retrieval to the CMS – we were able to put it into production within just a few months.
EBU T&I: How does recommendation technology tie in with your organization’s vision for the use of data?
Olle Zachrison: A long-term challenge for SR is to transform itself from an organization that serves large anonymous masses into a provider of relevant news content for individuals. Our platform competitors have fantastic recommendation engines that have set standards on the market, so it’s imperative that we advance our own capabilities. This allows us to keep up in terms of serving the audience, but also to maintain our independence and the public service values that are the foundation of our journalism. We certainly think that Peach helps us on this journey, especially as the project is so responsive to our ever-shifting needs and product plans.
EBU T&I: What are your future plans?
Olle Zachrison: In the area of news, we at SR are currently constructing our own version of a 'public service algorithm'. This news algorithm will be a core component when we start personalizing our news products to a larger extent than we do today. We want it to not just serve the individual consumer with an attractive and personalized mix of topics, locations and genres, but also to ensure that news items with great public service value have a prominent place in the listings. The knowledge and contacts that we get through our participation in the Peach project have already helped us to develop our thinking and to set up the project.
Jörgen Bang: To achieve these goals we are going to add more smartness to the algorithm that finds related news content. And we want to look at how we can bring this more personalized news offering directly to our audience. Our personalization is currently for on-demand and podcast content, and we want to explore how we can use Peach to help serve each listener their most relevant news stories in other ways.