Phailin_annotated

This data contains 1,500 manually annotated tweet records from the Cyclone Phailin dataset, representing a multilingual (predominantly English) subset of Twitter posts collected during 9–13 October 2013.
Each record corresponds to a single tweet and includes anonymized metadata, derived variables, and human-assigned categorical labels.

This dataset is designed for supervised machine learning and communication analysis, particularly for identifying informative content, tweet categories, and user types during natural disasters.
All direct personal identifiers (e.g., screen_name, username, user_id, profile text, precise location) have been removed or pseudonymized (user_anon_id).
Mentions and URLs in tweet text were stripped, while hashtags were retained; a cleaned tokenized version of the tweet is provided as tokenized_text.