Please wait for chart to load...

How is this computed?

Millions of Twitter users publicly post short text messages (tweets). We first find all tweets with emoticons (smiley faces), and use these as indicators of happy and sad emotions. These example tweets (several million of them) then allow a computer to learn what words and phrases typically indicate one mood or the other. After learning these phrases, the computer can then label all tweets with an emotion, resulting in a final percentage.

Why?

Ongoing research is investigating how public mood relates to seemingly unrelated events such as financial market movements, employment, and even political views. Our initial work found a surprising correlation between mood and presidential approval polls. The extent to which Twitter posts actually correspond to an average mood remains an open question. We offer this data freely to researchers and hobbyists alike.

Published Work

Learning for Microblogs with Distant Supervision: Political Forecasting with Twitter.
Micol Marchetti-Bowick and Nathanael Chambers.
In Proceedings of the European Association for Computational Linguistics. Avignon, France. 2012.
[PDF download]

Quick Facts

1. Average number of tweets per day: 3 million
2. Total number of tweets processed: 580 million
3. Data size: 72 gigabytes
4. Number of tweets with the basic smiley emoticon: 25 million :)
5. Happiness and sad lines correlate with Gallup's Presidential Approval polls (over two separate 7-month periods).

Research Data

Download the above data [.csv format]

Download hand-labeled tweets for the political domain