Global evidence of expressed sentiment alterations during the COVID-19 pandemic

Replication materials for J Wang#, Y Fan#, J Palacios, Y Chai, N Jeanrenaud, N Obradovich, C Zhou*, S Zheng* (2021).

The materials in this repository allow users to reproduce the data analysis and figures appearing in the paper.

If you have questions or suggestions, please contact Jianghao Wang at wangjh@mit.edu | wangjh@lreis.ac.cn

Computational requirement

• R 4.0+
• Python 3.7-3.9
• Stata 14.0+

Organization of repository

• input: all the necessary input data
• figures: the main text final figures
• script:
  • 01_sentiment/ : sentiment imputation, see the repository: https://github.com/MIT-SUL-Team/global-sentiment
    • data: the traning and labeled_data for the global sentiment imputation
    • dict/sentiment_dicts: the emoji, hedonometer, and LIWC dictionaries
    • models: the multilingual data for the sentiment
    • notebooks: sentiment clf evaluator.ipynb
    • output
    • report
    • src: main model and sentiment imputation folders
      • main_geography_imputer.py
      • main_sentiment_aggregator.py
      • main_sentiment_imputer.py
      • setup_emb_clf.py
      • setup_liwc.py
      • utils: functions used for the sentiment imputation
        • aggregation_utils.py
        • data_read_in.py
        • dict_sentiment_imputer.py
        • emb_clf_setup_utils.py
        • emb_sentiment_imputer.py
  • 02_visual/: exploration analysis, see details in figures.
  • 03_sentiment_recovery/: This section reproduce the result of Expressed sentiment alterations during COVID-19 pandemic： the first measure--recovery half-life.
  • 04_sentiment_shock_and_lockdown_effect/: This section reproduce the result of Expressed sentiment alterations during COVID-19 pandemic： the second measure--sentiment drop and the results of Impacts of lockdowns on expressed sentiment