Tracing Political Positioning of Dutch Newspapers

Abstract

Newspapers write for a particular readership and from a certain ideological or political perspective. This paper applies various natural language processing methods to newspaper articles to analyse to which extent the ideological positioning of newspapers is reflected in their writing. Political bias is illustrated in terms of coverage bias and agenda setting by means of metrics, LDA topic modelling and word embeddings. Furthermore, article source discrimination is analysed by applying various classification models. Finally, the use of generative models (GPT-2) is explored for this purpose. These analyses showed several indications of political tendencies: disproportionate coverage of certain politicians and parties, limited overlap of political discourse, classifiable article source and divergence of generated text thematically and in terms of sentiment. Therefore, reading a newspaper requires a critical attitude which considers the intricate political tendencies of the source.