Paper Overview
Field: NLP Authors: Flávio Soriano, Victoria F. Mello, Pedro B. Rigueira, Gisele L. Pappa, Wagner Meira, Ana Paula Couto da Silva, Jussara M. Almeida arXiv: 2604.21897
Abstract (translated)
Analyses of legislative behavior typically rely on voting records, overlooking the rich semantic and rhetorical content of political speeches. In this paper, the authors pose three complementary questions about parliamentary discourse: how things are said, what is being said, and who speaks in a discourse-similar manner. To answer these questions, they introduce a scalable and generalizable computational framework that combines diachronic stylometric analysis, contextual topic modeling, and semantic clustering of legislator speeches.
The framework is applied in a large-scale case study of the Brazilian Chamber of Deputies, using a corpus of over 450,000 speeches delivered between 2003 and 2025. The results reveal:
- A long-term stylistic shift toward shorter, more direct speeches.
- Sharp redirection of the legislative agenda during national crises.
- Fine-grained discourse alignment maps in which regional and gender identities are often more salient than formal party affiliation.
- Paper: https://arxiv.org/abs/2604.21897
More broadly, the work provides a robust methodology for analyzing parliamentary discourse as a multidimensional phenomenon that complements traditional voting-based approaches.