Democratic speech in the UNSC - Analyzing the rhetoric of speeches with BERT based on the Democratic Peace Theory

Conflicts between countries in the present day mostly do not escalate into direct wars but are rather resolved through diplomatic disputes. One of the most important arenas for these disputes is the United Nations Security Council where carefully drafted speeches are used to express one country's stance in foreign relations. In this study, we will explore whether the implications of the democratic peace theory - namely democracies being less prone to go to war with each other and thus being more peaceful - can be applied to rhetorical disputes as a new form of conflict behavior. Various research has already shown that this theoretical assumption has practical implications and is regarded by policymakers as a law-like rule, regulating relations between states (Ish-Shalom 2008). Therefore, we investigate the hostility in UN Security Council speeches by using the pre-trained language model BERT to classify speeches in the UNSC dataset from Schoenfeld et al. (2019) on a sentence level as hostile and non-hostile. We then use named entity recognition to test whether state representatives use less hostile language when addressing countries with the same regime type compared to countries with different regime types.