Link: https://zenodo.org/record/1048820#.W_BHg-JReMq
Paper link: Characterizing Political Fake News in Twitter by its Meta-Data
Task:
- binary classification - according to is_fake_news attribute
- multiclass classification - according to fake_news_category attribute
Collection of tweets related to the 2016 US election that went viral during the election day (Nov 8th). Viral tweets are those that achieved the 1000-retweet threshold duing the collection period.
We queried Twitter's streaming API using the hashtags #MyVote2016, #ElectionDay, #electionnight, and the user handles @realDonaldTrump and @HillaryClinton.
Tweets have been labelled as containing fake news or not by one expert. A fake news is one the following:
- Serious fabrication
- Large-scale hoaxes
- Jokes taken at face value
- Slanted reporting of real facts
- Stories where the 'truth' is contentious
- is_fake_news - whether message is fake or not
- fake_news_category - concrete category of fake news
- tweet_id - id of tweet
- created_at - date of tweet creation
- retweet_count - how many times was tweet retweeted
- text - text of the tweet
- user_screen_name - name of user that tweeted
- user_verified - whether is user verified or not
- user_friends_count - number of user's friends
- user_followers_count - number of user's followers
- user_favourites_count - number of user's favourites
- tweet_source - link to tweet
- geo_coordinates - geo coordinates
- num_hashtags - number of hashtags in tweet
- num_mentions - number of mentions in tweet
- num_urls - number of urls in tweet
- num_media - number of media in tweet