Analysing Calls to Order in German Parliamentary Debates

A team of researchers including Nina Smirnova, Daniel Dan, and Philipp Mayr has published a novel study applying natural language processing (NLP) to political science. They developed a rule-based method to automatically detect and annotate formal 'Calls to Order' (CtOs) in the German Bundestag, creating the first comprehensive dataset of parliamentary incivility spanning 72 years of debates. This dataset, accepted for the PoliticalNLP 2026 workshop, provides a quantitative lens on political conflict and norm violations that have previously been studied only anecdotally.

The analysis reveals systematic biases in how parliamentary rules are enforced. The data shows that male members of parliament and those belonging to opposition parties receive significantly more formal reprimands than their female and coalition-party counterparts. The most frequent trigger for a Call to Order was a personal insult directed at an individual. Furthermore, the research indicates that despite formal regulations, the issuance of CtOs is influenced by subjective factors, including the presiding session president and the dynamics of the debate itself, with debates on governmental affairs and presidential actions seeing the highest concentration of triggers.

This work establishes a new benchmark for computational political science, moving beyond simple sentiment analysis to a structured classification of specific rule-breaking speech acts. The publicly released dataset and methodology allow other researchers to study political polarization, institutional conflict, and the enforcement of deliberative norms over decades with unprecedented precision.