This project is part of my dissertation research, investigating how corruption scandals influence politicians rhetoric and policy priorities. Using over 13,000 campaign manifestos from Brazilian mayoral elections, I applied text classification techniques to determine whether corruption in specific policy areas (e.g., health) leads politicians to increase or decrease their focus on those topics in their proposals. To tackle this challenge, I manually labeled a subset of the data to train machine learning models, leveraging external APIs to enhance classification accuracy. This work demonstrates my ability to handle unstructured text data, apply machine learning, and develop creative solutions with limited resources.
GitHub Repository | Slides | Paper