Brazilian parties reflect senators opinions?
Clusters, Senators, Parties, Ideology, Natural Language Processing, Speeches, Federal Senate, Party cohesion.
This work aims to verify whether senators from the same party have similar ideological points of view. We measure the senators’ ideological points using their speechs and natural language processing techniques. We build our method based on three steps. First, we clean the speechs using nature language pre-processing techniques. Second, we split the senators into clusters according to the similarity of their speeches. Third, we compare the composition of the endogenously formed clusters with the composition of the parties. Our dataset of speechs come from the 51th to the 55th Brazilian senate legislature. We find that senators from the same party tend to have similar speeches. This also occurs between parties of the same ideology. Furthermore, we characterize each cluster with its most relevant words. This kind of characterization allows the identification of the position of the party as left, center or right..