data

Introduction

Here we release the data.

Facebook candidate posts

See /facebook_candidate. We share the IDs of the posts in CSV files with the following schema:

Column	Note
post_id	ID of the post
post_url	URL of the post

Facebook and Instagram posts based on keywords

See /facebook_instagram_keyword. We share the IDs of the posts in CSV files with the following schema:

Column	Note
post_id	ID of the post
platform	facebook or instagram
post_url	URL of the post
relevant*	1=related to midterm; 0=irrelevant

Post relevance label: see the paper for details. The procedure is not perfect, use with discretion.

Reddit submissions and comments based on keywords

See /reddit_keyword.

We share the data from each day in json files. Each line of the file is an object. We add a relevant label to each object.

Candidate tweets

See /twitter_candidate. We share the IDs of the tweets in CSV files with the following schema:

Column	Note
tid	ID of the tweet

Tweets based on keywords

See /twitter_keyword. We share the IDs of the tweets in CSV files with the following schema:

Column	Note
tid	ID of the tweet
relevant*	1=related to midterm; 0=irrelevant

Post relevance label: see the paper for details. The procedure is not perfect, use with discretion.

Facebook ads

See facebook_ads_ids.csv.gz . We share the IDs of the ads in CSV files with the following schema:

Column	Note
id	ID of the ad

4chan data

Since the 4chan data is too large for GitHub, we host it on zenodo.

For the archive files, each line is a thread in json format. For the snapshot files, each line is a json object containing the snapshots of 4chan's catalog endpoint.

Note: we did not perform any keyword matching on the 4chan data so not all content is related to the 2022 US midterm elections.