Detecting Mental Health Disorders Using NLP

The concept of mental fitness pertains to the state of possessing a sound and adaptable mind, characterized by good health and resilience. It encompasses the capacity to navigate challenges and setbacks in a constructive and flexible manner. Consequently, the cultivation and sustenance of mental fitness contribute significantly to the improvement of individual well-being and overall quality of life. In 2019, the WHO observation showed that close to a billion people around the world were living with a mental disorder especially depression and anxiety being the most common. There are various NLP projects which deal with classifying a specific mental health disorder based on analysis of text or speech but very few deal with multi-level classification. To address this we are doing a single model with multilevel mental health disorder classification-combining the text data from many data sets of individual classes. To address this we are creating a composite framework with multilevel mental health disorder classification-combining the text data of different classes from a dataset of reddit posts. Our experiments have shown that fine-tuned domain-specific transformer models and hybrid models combining transformer models and recurrent neural network models like LSTM and Bi-LSTM can detect mental health disorder with an fl-score of 88 % and above.