The seven emotion demonstration below is part of Existor’s machine learning work. If you are interested in this seven emotion model, please contact us.
About the model
This model uses a word-based RNN to guess the emotion/feeling felt by the author of a piece of text. It gives a percentage score for flirty, angry, sad, disgust, surprise, laughter and happy. These represent the types of emotions that Cleverbot users express when interacting with our chatbot.
Our model analyses how the user was feeling at the time of writing. It works on single words like “good” or whole sentences “I feel good.” This is different from the popular sentiment analysis tools developed from hand-labelled movie reviews. For example a 5-star review of “That was such a sad story.” would be considered a very positive review, even if the user felt very sad when writing it. The data we use comes from Cleverbot conversations where users essentially labelled their own statements.
The sequence of words is important in any recurrent neural network. So “You kissed my friend.” and “My friend kissed you.” will produce different results, and someone who types “You kissed me.” is a happier than just “You kissed.”.
