Application of natural language processing to the electoral manifestos of social democratic parties in Central Eastern European countries

Abstract

The paper examines electoral manifestos of social democratic parties in Visegrad countries through a computer-driven method of content analysis. The analysis focuses on a sentence sentiment (how parties talk about their priorities) and a keyword examination (what parties mostly talk about). Such analysis is not widespread in the academic literature related to Visegrad countries, where a human-based approach to content analysis of electoral manifestos prevails. The paper, however, does not aim to nullify the findings of the previous research, but aims to improve the understanding of the content of manifestos through a different research method. The paper's findings have confirmed the pro-European support of social democratic parties as well as the focus on predominantly socio-economic issues in manifestos. Additionally, the paper explores the effect of incumbency on the manifesto's content. The incumbency seems to have an impact on the manifesto sentiment, but not on keywords.