PsychBERT Fine-Tuned on Simple Multi-class Problem

This is a version of https://huggingface.co/mnaylor/psychbert-cased which was fine-tuned to illustrate performance on a multi-class classification problem involving the detection of different types of language relating to mental health. The classes are as follows:

0: Negative / unrelated to mental health
1: Mental illnesses
2: Anxiety
3: Depression
4: Social anxiety
5: Loneliness

The dataset for this model was taken from Reddit and Twitter, and labels were assigned based on the post appearing in certain subreddits or containing certain hashtags. For more information, see the PsychBERT paper.