Algorithm learns emotions to recognize in text using emoji
American scientists have developed an algorithm that sarcasm, and emotions in text can recognize by taking into account the use of emoji.
At MIT Technology Review tell the researchers that they had a method that wanted to develop for racist Twitter messages to recognize.
That, however, proved to be difficult without knowing when Twitter users a sarcastic message placed. Therefore, to troops them their ‘deep learning’algorithm to emotions in tweets to recognize, and take account of the emoji that people use.
“Because we have no intonation of our voice or body language can use context to what we say, we use online emoji,” says MIT researcher Iyad Rahwan. “The neural network has a connection between a certain kind of language, and emoji.”
The algorithm DeepMoji ‘learned’ of 1.2 billion tweets in which at least one of 64 most popular emoji was used. Using an existing dataset, the algorithm can also be trained to sarcasm to recognize it.
According to the researchers, performing their algorithm is much better at recognizing emotions than existing techniques. In a test with human subjects even proved that the algorithm more often sarcasm can correctly recognize (82 percent) than men (76 percent).
The algorithm is made public, so that other researchers start working with them. On a special site, people may have their own tweets to be analyzed, and by means of emoji to pass what emotions are associated with it. As the algorithm is getting better and better.