During my master thesis, I tried to approach depression detection on Twitter with deep learning techniques. I had mainly three different problems to solve :
- Mutli-classification of emotions ( happy, fun, love, neutral, sad, worried)
- Emotions Prediction
- Depression Detection
The simple idea behind is the following : the model takes the last 30 messages of the user, determines their emotions, observes their evolution through and says whether the user is depressed or not.
My thesis for even more details : Deep Learning for Emotions Analysis.
I am currently creating an app, Spleenr, to give access to everyone to my models.