MediaDiet

Important results from the project

An AI based recommendation engine for local newspaper news was developed, used in production for ten months, and results analyzed in accordance with the purpose of the project: to provide a more relevant news flow and increased engagement for news readers, supporting in-depth reading and avoiding filter bubble. Furthermore, this has led to increased knowledge about users´ news consumption and behavior patterns, filter bubbles and navigation flows, and how these are affected by recommendations.

Expected long term effects

The expected results associated with the project work package have mostly been achieved. The developed recommendation engine and the knowledge of user behavior and filter bubbles achieved within the project have created new opportunities for user-friendly news feeds that encourage both broad and deep news consumption. The recommendations are, however, currently performing lower than expected, but further analysis and development is ongoing with the aim of improving hit accuracy.

Approach and implementation

The project has largely been implemented in line with the project plan except that the closing date was postponed by two months at the request of iMatrics. During the course of the project, we have had regular physical meetings as well as meetings through digital tools to discuss work progression and plan upcoming steps.