Opportunity to improve content creation with additional insights
Alma Media had been interested in finding ways to improve operations and outputs in the newsroom. The traditional creative process asks journalists to intuitively predict the success of content, understand the particular tastes of readership and know the perfect timing for a news piece.
Yet, there’s little to no data available to help journalists come up with successful future content. Alma Media collaborated with Fourkind to design a predictive classification algorithm that uses data to anticipate the success of editorial content.
Deep learning to switch from intuition to data-driven journalism
The Media Industry Research Foundation of Finland funded the pre-study phase, where Fourkind conducted interviews with decision-makers of global media companies to understand the predictive models suitable for journalistic purposes. After interviews and workshops to rank areas for potential growth drivers, we determined a proof of concept. Editors were also involved to help ensure that the new tool supported the journalists and reduced the risk of resistance to adoption. Every department had the opportunity to voice their opinion and the whole team worked towards a common goal of creating best-in-class content for readers.
The chosen concept was a predictive tool for newsroom content creation for Aamulehti, the second largest newspaper by circulation in Finland. We used state-of-the-art natural language processing models and a million Alma Media articles to teach the algorithm how the Finnish language is used in newspapers. On top of the language model, we built a classification model that uses existing user data to predict the success of future content.
Machine learning changes the media business
The tailor-made, data-based classification tool was integrated into Alma Media’s content creation software, to help journalists get a prediction of the article’s future success with just one click. The editors and journalists themselves still have the final call, but the classification algorithm provides insight to write news stories that will reach more people.
Since the start of the project in early 2019, the collaboration has been marked by mutual respect and trust among the editorial department. The machine learning project has helped journalists improve articles, and eliminate ineffective content and operations have improved for the newsroom with new in-house data capabilities, an enhanced data-based editorial culture and improved agility for future processes.
We are pleased with the predictive tool’s achieved accuracy since it’s critical that the editorial can be confident with the results.
The model was later deployed in Alma Media’s other publication, Iltalehti. As Alma Media’s needs develop, the tool can be expanded to include upgrades.
