An AI-Driven Hybrid Approach for Detecting Mental Health Indicators in Multilingual Indian Social Media: Data Acquisition and Analytical Frameworks

Social networks have become the leading platform for human expression and have therefore been helpful in the early detection of psychological distress. This paper presents a multilingual framework to harvest and examine social media posts in a variety of regional Indian languages, such as Tamil, Telugu, Kannada, Malayalam, and English, to identify signs of different mental health disorders, such as depression, anxiety, and stress. The proposed method combines cutting-edge natural language processing and deep learning approaches, with special attention to exploiting Transformer models in conjunction with psychological lexicon-features, to disentangle complex linguistic and emotional patterns in multilingual text. A high-quality dataset, annotated by mental health experts to capture a variety of mental health signs, served as the basis for model training. The preprocessing pipeline addresses the challenges of multilingual, code-mixed, and transliterated text to provide uniform data quality across languages. The evaluation results show that the proposed hybrid model performs better than classical sentiment analysis approaches, achieving higher accuracy in identifying a variety of mental health signs. By identifying subtle emotional and linguistic signals in social media posts, this work illustrates the viability of scalable, real-world mental health monitoring across India's linguistic diversity. Future work includes the extension of language coverage, model interpretability through explainable AI, and real-time application to assist mental health clinicians in early detection and intervention.