DistilBERT-based Text Classification for Automated Diagnosis of Mental Health Conditions

Mental health disorders present a substantial global health challenge, underscoring the critical importance of timely diagnosis and intervention for effective treatment. However, the proliferation of textual data within online mental health communities has underscored the need for automated methods that can classify and diagnose individuals based on their expressed thoughts and emotions. Existing diagnostic practices for these conditions primarily rely on labour-intensive and time-consuming clinical assessments and interviews conducted by healthcare professionals. Notably, a key challenge in this context is the absence of robust physiological indicators for mental disorders. This paper addresses this challenge by proposing a DistilBERT-based text classification approach for the automated diagnosis of mental health conditions. In addition, recognizing the emerging landscape of interdisciplinary exploration into the gut brain axis, we acknowledge the potential role of gut bacteria, the microbiome, and microorganisms in influencing mental health. Our research specifically focuses on three distinct mental health disorder conditions: anxiety, borderline personality disorder (BPD), and autism. To ensure the robustness of our approach, we curated a balanced dataset, comprising 500 samples for each of these three classes. This diligent effort yielded a noteworthy achievement, with our model attaining a remarkable accuracy rate of 96%.