Measuring semantic polarisation of four Swiss media brands with topic modelling

Polarisation increases content views, perceived importance of the content, and engagement. However, it can also harm customer relationships and brand reputation. Polarisation can be affective or semantic in nature[1]. The former has already been examined computationally, but not the latter. Semantic polarisation is examined in 2,337 brand-related articles from two polarising and two non-polarising Swiss media brands with topic modelling. Words referring to the two polarising brands can be found on both levels, the topic model and topic cluster. This does not hold true for non-polarising brands. Topics that semantically polarise pertain to i) the Swiss Radio and Television, ii) Facebook and Instagram, and iii) Putin, Russians and Sochi. This paper aims to contribute to polarisation research suggesting the computational examination of semantic polarisation. The findings can be applied in media, communication and marketing practice.