metadata

license: ecl-2.0
task_categories:
  - text-classification
language:
  - en
  - es
tags:
  - public
  - text
  - education
  - sentiment-analysis
pretty_name: TweetFeels 1m4
size_categories:
  - 100K<n<1M

TweetFeels 1m4

An 1-million-tweet sentiment corpus harvested from Twitter and annotated with four fine-grained categories: positive, negative, uncertainty, and litigious. Each record carries three clean, tab-separated fields:

Language – ISO-639 code of the tweet’s detected language
Text – the full tweet text, preserved with original casing and emojis
Label – one of {positive, negative, uncertainty, litigious} determined by an automated labelling pipeline

The dataset covers multiple languages and informal, real-time expressions typical of Twitter, offering a sizeable, ready-to-use resource for multi-class sentiment and legal-tone modelling.

Acknowledgements: The dataset is hosted on Kaggle—Sentiment Dataset with 1 Million Tweets