This is the dataset containing tweets for POI type prediction.  If you use this data please cite the paper: D. Sánchez Villegas, D. Preoţiuc-Pietro and N. Aletras (2020). Point-of-Interest Type Inference from Social Media Text. In AACL.

The dataset is divided into train/dev/test sets. Each row represents one tweet, containing {tweet_id, user_id, category_name, time, place_id}.

tweet_id: the id of the given tweet
user_id: the Twitter ID of the user
category_name: the POI category
time: the time when the tweet was posted (YYYY-MM-DD HH:MM:SS)
place_id: the Twitter ID of the place

The tweet IDs can be used to retrieve the original tweet using the Twitter API, alongside any other information you may require. Note that some of the tweets may have been deleted or set to private.
