Repository for the final project of the seminar “Social Media Data Analysis.” This project examines how German politicians change the sentiment of their social media posts over the course of the election cycle. In particular, it examines whether and how members of parliament communicate differently during election campaigns. This project is a replication of the study Modulation of Democracy: Partisan Communication During and After Election Campaigns (Castanho Silva et al., 2024).
Contains all retrieved, preprocessed and labelled data.
- clean_data All extracted and cleaned Bluesky posts, cleaned MP-level dataframes and manually labelled sample posts
- clean_labelled_data All classified posts and grouped per weekly intervals
- raw_data The raw MP data for both legislatures with their potential handles
Contains the code for retrieving data on Bundestag MPs as well as the retrieval of all of their Bluesky posts.
Note: An app password is needed for post retrieval which is not shared in this repo.
Contains all exported plots as png and tables as tex files for the report.
Contains code that performs preliminary data analysis both on both MP and post-level data.
Contains both the validation of different sentiment classifiers as well as the final classification script.
Contains the code for the main analysis and a robustness check.
Contains the main report and the initial project presentation.