ogencoglu / causal_twitter_modeling_covid19

Source code for the paper "Causal Modeling of Twitter Activity during COVID-19". Computation, 2020.

Implementation of Causal Modeling of Twitter Activity during COVID-19 - Gencoglu O. & Gruber M. (2020) ==================== This repository provides the full implementation. Requires python 3.7.

Main Idea

Distinguishing events that "correlate" with public attention and sentiment change from events that "cause" public attention and sentiment change during COVID-19 pandemic

Quick Glance at Findings ===================

Reproduction of Results ==================== 1 - Get the data --------------See directory_info in the data directory for the expected files. Template of tweets.csv is provided.

2 - Run causal_inference.ipynb -------------------------------See source directory.

Relevant configurations are defined in configs.py, e.g.:

--start_date '2020-01-22' --end_date '2020-03-18' --sentiment_model 'distilbert-base-uncased-finetuned-sst-2-english' --percentiles [75]

source directory tree:

├── causal_inference.ipynb
├── configs.py
├── data_utils.py
├── eval_utils.py
├── feature_extraction.py
├── inference.py
├── sentiment.py
├── train.py
└── train_utils.py

Cite ====================

@article{gencoglu2020causal,
  title={Causal Modeling of Twitter Activity during COVID-19},
  author={Gencoglu, Oguzhan and Gruber, Mathias},
  journal={Computation},
  volume={8},
  number={4},
  pages={85},
  year={2020},
  doi={10.3390/computation8040085}
}

or

Gencoglu, Oguzhan, and Gruber, Mathias. "Causal Modeling of Twitter Activity during COVID-19." Computation. 2020; 8(4):85.