NewsDNA: an algorithmic approach to news diversity
Camiel Colruyt

In the modern media landscape, news recommendation algorithms increasingly shape the consumption of news for a significant part of the audience. This algorithmization leads to less diversity in the views that audiences are exposed to: the so-called filter bubble (Pariser 2011). NewsDNA is an interdisciplinary research project which explores news diversity through the lenses of language technology, recommendation systems, communication sciences and law. Its aim is to develop an algorithm that uses news diversity as a driver for personalized news recommendation.
This talk presents the work LT3 has performed on the NewsDNA project since its start in March 2018. When different newspapers report on the same news events, they may do so with a different framing or perspective, reflecting the paper’s viewpoint. We aim to discover whether it is possible to detect these perspectives computationally. This problem is related to stance detection, where the relative stance of documents towards a given theme is predicted (Mohammad et al. 2016). The theme is usually a topic or entity; in our case, we target news events.
To acquire the targets, we will perform event extraction: automatically discovering what news events are brought up in a newspaper article. We present an annotation system for news text inspired by the ACE framework (Doddington et al. 2004) and we highlight the challenges of this design exercise. This annotation system will be applied to a corpus of articles from Flemish newspapers to create a new Dutch-language event corpus.