Viral Tweets Labeled for Fake News During the 2016 US Election

Julio Amador from Imperial College London collected tweets that went viral on the 2016 US election day. The collection includes tweets using specific hashtags and from candidate accounts, each labeled by an expert as containing fake news or not. Fake news labels cover categories like serious fabrication, large-scale hoaxes, jokes taken at face value, slanted reporting, and contentious stories.

Use Cases

• Train classifiers to detect fake news based on expert-labeled tweets.
• Analyze the characteristics and spread of viral political misinformation.
• Study the relationship between specific hashtags, user mentions, and the presence of fake news.
• Investigate the types of fabricated or slanted reporting that gained traction during a major election.

Strengths

• Tweets are labeled by an expert for fake news across five defined categories.
• Data collection targeted specific, relevant hashtags (#MyVote2016, #ElectionDay, #electionnight) and user handles (@realDonaldTrump, @HillaryClinton).
• Focuses on viral content, defined as tweets achieving a 1000-retweet threshold.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect temporal and platform bias inherent to the Twitter API and the single election day focus.

Provenance

Source

Imperial College London

Collection Method

Collected via Twitter's streaming API using specific hashtags and user handles, with expert labeling.

Time Range

Election day, November 8th, 2016

Geography

United States