International Journal of Scientific Research and Engineering Development

Mental health disorders, including depression, anxiety, and suicidal ideation, represent a growing global crisis, yet early detection and continuous monitoring remain inadequate. Conventional assessment methods relying on clinical interviews and self-report questionnaires are episodic and fail to capture the ongoing fluctuations in an individual's psychological state. This paper presents MentalBERT-Track, an AI-powered system that performs real-time sentiment analysis on usergenerated text to detect, classify, and longitudinally track mental health risk. The system leverages transformer-based Natural Language Processing (NLP) architectures—specifically a domain-fine-tuned BERT variant—to extract sentiment polarity, emotional intensity, and suicide-related linguistic markers. A temporal trend module aggregates daily sentiment scores to identify deteriorating emotional trajectories and trigger timely alerts. Evaluated on a corpus of 12,000 annotated social-media posts and online forum entries, MentalBERT-Track achieves 94.7% classification accuracy, an F1-score of 93.8%, precision of 94.1%, and recall of 93.5%. These results surpass baseline SVM, LSTM, and standard BERT configurations, demonstrating that domain-adapted transformer models with temporal modelling significantly enhance mental health risk stratification. The system operates entirely on text, requiring no clinical data or physiological sensors, making it scalable and privacy-preserving for deployment in digital mental health contexts.

Nidhi Sharma, Monika Sharma, Pushpendra Kumar Dwivedi,"An AI Approach to Mental Health Monitoring Using Sentiment Analysis of User Data" International Journal of Scientific Research and Engineering Development, V9(3): Page(642-651) May-June 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.