In this repository we provide dataset used and produced in:

Temporal and geographic analysis of the Hydroxychloroquine controversy in the French Twittosphere Faccin, Emilien, Gargiulo arXiv:2304.14075 (2023).

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 refer to this dataset as:

DataCovHCQ Mauro Faccin Zenodo (2023)

The file data/african_tweets_YYYY-MM.csv.gz contains the list of tweets and retweets IDs used in the above publication, split by month. The file ./data/african_users_extended.csv.gz contains the list of users IDs as well as their country (in the geo_coding_extended column one can find the full geographic tagging used).

This dataset uses the same keywords as in DataCovVac for Twitter streaming and search APIs. The dataset ./data/keywords.json is divided into three sets that corresponds to the three tweet datasets (DataVac, DataCov, DataHC).

The dataset format is as follows:

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

In the covhcq-code folder we provide a number of python scripts that extract the data needed to plot. The code requires tweets and retweets to be hydrated. Files containing the tweets should be named ./data/african_tweets_YYYY-MM.csv.gz and have the following columns:

id,time,created_at,text,from_user_name,from_user_id,to_user_id,quoted_id,quoted_user_id,mentioned_user_ids,location,links,hashtags

Files containing retweets should be named ./data/african_retweets_YYYY-MM.csv.gz and have the following columns:

id,retweeted_id,time,from_user_id,created_at

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

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