Mental Health Prediction through Text Chat Conversations

This study investigated the viability of classifying mental health conditions from text chat conversations using deep learning (DL) models. A novel dataset of real-time users was acquired from the Lyf Support app, which was labeled by psychologists to conduct the experiments for users' mental health prediction. The findings suggest that DL models could be used to identify mental health conditions from text chat conversations. According to evaluated results, the bidirectional gated recurrent unit (BiGRU) outperformed other models by achieving over 83% accuracy. The dataset employed in this study may be crucial for future research on online social media users' mental health.