A Binary Question Answering System for Diagnosing Mental Health Syndromes powered by Large Language Model with Custom-Built Dataset

With telemedicine and remote patient monitoring, the need for accessible mental health resources has become increasingly apparent. This work presents a binary question-answering system tailored for individuals experiencing depression, utilizing Large Language Models (LLMs) designed to handle the complexities of mental health inquiries. By examining existing LLMs, a system enhanced by prompt engineering on pre-trained models was developed to meet the particular necessities of those seeking help for depression. Utilizing depression inventories and dictionaries, a detailed dataset of depression symptoms was compiled, creating precise prompts to elicit comprehensive and relevant responses from the LLMs. The system was tested using zero-shot, one-shot, and few-shot learning techniques, with results compared against human-generated responses to assess efficacy. This study confirms the flexibility and accuracy of LLMs in managing sensitive mental health conversations and serves as a foundation for future developments. Further enhancement of the system is proposed through specialized fine-tuning of LLMs with a dedicated corpus of depression-related prompts. The goal is to deliver a scalable, empathetic tool that provides timely and effective support for individuals battling mental health issues, leveraging cutting-edge AI technology to address a critical gap in mental health care.