We use millions of texts on Twitter containing emojis for training a deep learning model that understands many nuances of how language is used to express emotions. For instance, it does well at capturing sarcasm and slang. We beat state-of-the-art algorithms across many benchmarks datasets. See our paper, blog post or FAQ for more details.

Our pretrained DeepMoji model is released for anyone to use (incl. companies), and implementations are available for the Keras and pyTorch frameworks.

DeepMoji is based on research done by Bjarke Felbo, Alan Mislove, Anders Søgaard, Iyad Rahwan and Sune Lehmann. The research to further improve our understanding of emotional content in text is carried out by Bjarke Felbo, Nick Obradovich, Holly Shablack, Kristen Lindquist and Iyad Rahwan. The website was designed by Max Lever.

Disclaimer: Note that the model has learned about language from the raw, uncurated expressions of individuals on social media. We do not endorse in any way the emotional interpretation that the model has of any particular content.

Contact us

Selected Media

MIT Technology Review, BBC, Newsweek, Business Insider, The Telegraph, The Register, Huffington Post (FR), Numerama (FR)

Contact Us

Send us any comments or concerns using the form below. Due to the amount of emails received it may take a while for us to respond. We appreciate your feedback.

If you feel you have been treated unfairly, or you have questions regarding your rights as a research subject, you may contact the Chairman of the MIT Committee on the Use of Humans as Experimental Subjects (COUHES) at couhes@mit.edu.