Negative Ties Highlight Hidden Extremes in Social Media Polarization

Open source data and code for the research paper:

E. Candellone,* S. A. Babul,* Ö. Togay, A. Bovet, and J. Garcia-Bernardo

Negative Ties Highlight Hidden Extremes in Social Media Polarization

Pre-print: https://arxiv.org/abs/2501.05590

Data:

*shared first authors

Contents of the repository

• /bertopic/: BERTopic intermediate results and model specifications
• /data/: CA and SHEEP embeddings and network files
• /figures/: paper figures
• /hsbm/: TM-hSBM intermediate results and model specifications
• /ideology_twitter/: validation with Twitter data and PoliticalWatch
• /notebooks/
  • 1_topic_modelling.ipynb: script to perform BERTopic and TM-hSBM topic modelling
  • 2_compare_hsbm_bert.ipynb: comparison of the two methods to have robust topics
  • 3a_create_attitudes.ipynb: create network embeddings using SHEEP and CA
  • 3b_sheep_null_model.ipynb: null model to compare SHEEP and CA
  • 4_figures_paper.ipynb: code to reproduce the figures of the paper
• /src/
  • create_snapshot.py: code to extract and clean data from scraped webpage.
  • topicmodelling.py helper functions for topic modelling.
  • meneame.py, s3_create_attitudes.py: helper functions for creating embeddings.

Instructions

1. Create conda environment:

conda env create -f polarization.yml
conda activate polarization

2. Run the notebooks

Contact

• Corresponding authors: Elena Candellone and Shazia Babul