Political discourse is constantly in flux: the key issues, the actors that define the discourseare changing from moment to moment. Quantitative approaches to discourse are an impor- tant methodological tool that can help researchers measure the structure of and changes in discourse. However, existing approaches are often time-consuming and not scalable to large datasets. In this thesis, I apply a fully automated approach to discourse analysis that combines Structural Topic Modeling and Network Analysis. I apply the model to a novel dataset of the Brexit negotiations in British parliament. I find that the method captures the state of relations between actors and the progress in the negotiation process.

Politics make us anxious. People experience anxiety – defined here as a fearful un- certainty about the future course of events – when thinking about political issues such as climate change and the economy, ahead of important elections, and even when hearing the voices of certain political figures. Yet, scholars have devoted little attention to the study of this critical emotion. We know that politics make us feel anxious, but what does the language of political anxiety look and sound like? By understanding the language of anxiety, we may someday be able to understand how political elites use anxiety as a tool to persuade colleagues and constituents alike. Using the Brexit negotiations (2016-2020) as a case study, I develop a methodological tool that measures anxiety and emotional intensity in elite political rhetoric using text and audio data. I develop a dictionary of anxiety that scores speeches based on the semantic similarity to a sample of highly anxious words. With this dictionary, I am able to study how anxiety varies with party affiliation and the topic of the speech. I then examine a large subset of the data in audio format using pitch as a proxy for emotional intensity. This novel approach combines the measure of anxiety in text and emotional intensity in audio, constructing a fuller picture of the speaker’s emotional state. The composite measure reveals how expressions of emotion differ according to the role of thespeaker in a meeting and their party affiliation.