metadata

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

TweetFeels 100k

A compact, automatically labeled Twitter-sentiment corpus of 101 160 tweets collected via the Twitter Search API in 2015. Polarity labels (0 = negative, 4 = positive) were assigned using distant supervision: tweets containing positive emoticons such as “:)” were deemed positive, while those with negative emoticons like “:(” were deemed negative.

Each record carries two tab-separated fields:

target – sentiment label (0 or 4)
tweet – full UTF-8 tweet text

The dataset focuses on short, informal language typical of Twitter and is sized for rapid experimentation, classroom use, or quick model prototyping in sentiment-analysis pipelines.

Acknowledgements: The dataset is hosted on Kaggle—Tweet Sentiment Dataset.