Assessing polarisation in brand-related comments on three Swiss online media portals with Natural Language Processing

Abstract

Polarisation increases content views, perceived importance of the content, and engagement. But it can also harm customer relationships. The degree of polarisation of four media brands is assessed in brand-related comments using sentiment and emotional intensity analysis and topic modelling. The degree of polarisation differs significantly by brand and topic. The three most polarising topics are law enforcement, Russia, and Corona transmissions. The three least polarising topics are broadcasting, entertainment and (Swiss) German language. Measuring polarisation helps to monitor brand performance.