When we met our client, a Belgian media group, they were at a crossroads. Their news business was undergoing rapid transformations and one of their challenges lied in poor digital adoption resulting in fewer subscribers, hence profit loss. In the meantime, their sports column was solid, driving most of their readership.

Personalised online news was most certainly the way to stand out. Yet, when such organisations embark upon implementing data-driven, algorithmic news personalisation services, they have to navigate a complex landscape. How to design it and implement it within their systems was a challenge we were eager to take.

When our client presented their challenges, it was very clear to us that their sports articles had to trigger readers’ loyalty.

So building algorithmic recommender systems on sports articles played a central role.

We kicked off this mission by  extracting topics from any article using natural language processing. Thanks  to frequently observed topics, our system was able to understand the meaning  of topics and sort them out.

Then, we created a news recommender  system based not only on users’ clicks and search but also on their individual favourite  topics to make recommendation even more accurate.