In this repository we provide dataset used and produced in:

Assessing the influence of French vaccine critics during the two first years of the COVID-19 pandemic Faccin, Gargiulo, Atlani-Duault and Ward PLOS ONE, 17(8) p.1-19 (2022).

The cited article contains a detailed description of the datasets' construction processes. If you use this in your project, please add a reference to the above article.

You may refere to this dataset as:

maurofaccin/DataCovVac Mauro Faccin Zenodo, (2023)

The file data/tids.XX.txt.gz contains the list of tweets and retweets IDs used in the above publication, splitted in chunks of 2M IDs.

The dataset contains 382 French websites and blogs that contains vaccine-critical postures (some of those URLs may not be working anymore), and 383 French websites of news media.

The dataset is saved in ./data/urls.json in the following format:

{
  "vaccine-critical: [...]",
  "news-media": [...]
}

This dataset includes all the keywords used to extract tweets from Twitter using its streaming and search APIs. The dataset ./data/keywords.json is divided into three sets that corresponds to the three tweet datasets (DataVac, DataCov, DataHC) mentioned in the paper above.

The dataset format is as follows:

{
    "DataVac": [...],
    "DataCov": [...],
    "DataHC": [...]
}

In the covvac-code folder we provide a number of python scripts that extract the data needed to plot. The code requires that the tweets (once hydrated) will be separated in daily chunks of:

YYYY-MM-DD-tweets.csv.gz
YYYY-MM-DD-retweets.csv.gz
YYYY-MM-DD-users.csv.gz

Users may be repeated on different days.

WARNING: script are provided as is, and they require user intervention in order to update paths and possibly other data.

In the covvac-plots folder one can find the python script used to reproduce the plots of the above paper.