Researchers at the Massachusetts Institute of Technology (MIT) have developed a deep-learning textual sentiment algorithm that was trained on emoji, and which can analyze tweets to pick up sarcasm and general emotional subtext.
"The neural network learned the connection between a certain kind of language and an emoji," says MIT professor Iyad Rahwan.
The DeepMoji algorithm was trained on 1.2 million tweets containing some combination of 64 popular emoji. The team trained the system to anticipate which emoji would be used with a particular message, based on whether the emoji reflected a particular emotion or sentiment. The researchers then taught DeepMoji to recognize sarcasm using an existing set of labeled examples.
Testing showed DeepMoji outperformed both other top sentiment-detecting algorithms and humans in the identification of sarcasm and other emotions on Twitter.
Experts say the research demonstrates that computers are gradually becoming more adept at sensing human emotion.
From Technology Review
View Full Article
Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA
No entries found