“Disgrace in the Polish Parliament”—analysis of parliamentary debates using generative models

Abstract

Over the last decade there has been a clear increase in extreme political rhetoric in the parliamentary debate in Poland. One aspect of the progressive radicalization of the language is overuse of words having a high negative emotional intensity to describe events in everyday politics, which leads to the degradation of their importance. As research shows, brutalization of the language of the debate has negative consequences at the levels of the political elite and the electorate, leading to reduced efficiency of the parliament, the exacerbation of conflicts between parties and among their supporters, and the political alienation of citizens. In this article, I present the possibilities of using the transcripts of parliamentary discourse to analyze the changes of the political discourse, including the algorithms used for searching text corpora and analysis of the latent topics. As an example, I use a set of parliamentary speeches from the years 1991-2016 related to the idea of shame, betrayal, disgrace and scandal. I base my analyses on generative topic modeling employing on the method of latent Dirichlet allocation and its extension—Structural Topic Model, both being unsupervised methods of extracting topics from large text corpora. The results of the analysis confirm the increase in usage of radical vocabulary in time, and also describe the characteristics of its users and the most common contexts the extreme vocabulary tend to appear (identity politics, control over the media, foreign policy).