Journalist and political science teacher, Fábio Vasconcelos, published a quick analysis on the blog Na Base dos Dados which we created on the parliamentarians’ speeches during the impeachment voting session in the Chamber of Deputies on 17th April.
For the curious, the database was the Chamber of Deputies’ PDF (yes, believe it or not). I tried to extract data from the speeches via R using the bRasilLegis package but I discovered that it will only be released between today (19th) and tomorrow. (Update: the speeches are now available).
After handling the PDF, I used Iramuteq. For those who do not know about it, Iramuteq is a “visual interface” to use R and produce textual analysis. It is the “offspring” of well-known French textual analysis software called Alceste. However, its techniques are strongly based on the textual analysis of French literature.
One of Iramuteq’s most interesting techniques is the Reinert method of collecting together word proximity. It allows you to find large thematic groups in the body of a text in a very simple way.
In this example, we found four large groups:
Termos

Each group’s position can be easily confirmed with regards to the neighbouring terms and the extent to which each group is close to another. Blue and green represent the Yes group; red the No group and there is a more residual group in grey (it has yes, no and also abstentions).
Iramuteq uses a dictionary to process the terms, so the term “querida” was dealt with as “querido”, as can be seen in the blue cloud. The States with a space in their name (Eg “São Paulo”) were replaced in the body by “são_paulo” for the software not to understand two distinct words.
Another really interesting option is factorial Correspondence Analysis (CA) of the illustrative variables. It allows the variables of each parliamentarian in relation to their speech (State, Vote and Party) to be analysed
 
The analysis was only of the parties in the following example.
Partidos
 
Some really interesting work such as Atlas Político and Radar Parlamentar use similar techniques to give the parties’ position in relation to the roll-call votes cast in the plenary session.  In this case, Iramuteq brought the parties together from the speech content.
(Updated)
An example which was not previously posted is the relationship between the States:
afcf_col
 

Postagens relacionadas:

  1. Conheça o Iramuteq – Ferramenta gratuita que permite transformar qualquer texto em dados valiosos.
  2. O que é uma pesquisa eleitoral? (Guia atualizado para 2022)
  3. Traduzindo a política em dados – Confira como transformamos discursos de parlamentares em informações sistemáticas e confiáveis
  4. No Brasil, Análise de Conteúdo é sinônimo de Bardin e porque isso é um problema
  5. Seminário do IBPAD discute Ciência de Dados como metodologia para estudos da sociedade e política

---
Para quem quer se aprofundar mais no uso de dados na política, o Ibpad lançou uma Formação completa com quatro cursos incríveis, confira:

Dados & Mapas – Identifique regiões de eleitores leais utilizando bases geolocalizadas
Pesquisas Eleitorais – Para tomar decisões corretas utilizando pesquisas de opinião
Pesquisa Qualitativa – Para entender como os eleitores pensam utilizando grupos focais
Inteligência de Dados em Mídias Sociais – Para quem quer ser relevante nas redes monitorando e analisando dados online