Predicting Mental Health Indicators with Enhanced BERT

Textual data from digital interactions contains vital early indicators of mental health challenges.

In this paper published at CML 2024, we introduced an enhanced BERT architecture fine-tuned specifically for sequence classification on nuanced psychological markers.

The model achieved significant accuracy improvements over standard baseline classifiers by capturing contextual semantics and subtle linguistic cues.

Our findings demonstrate the potential for automated early screening tools to assist mental health professionals.