IT 03 - WellMind: An AI-Driven Smart Diary for Mental Health Support

What is it?

WellMind is an innovative AI-powered application designed to address global mental health challenges by providing personalized and proactive mental health support. Built on state-of-the-art technologies like Retrieval-Augmented Generation (RAG), large language models (LLMs), and sentiment analysis, WellMind serves as a comprehensive smart diary platform that empowers individuals to understand and improve their mental well-being. With its intuitive interface, the application offers tools for mental health classification, personalized advice, wellness tracking, and seamless user management.

What does it do?

WellMind combines advanced AI techniques with user-friendly features to provide the following core functionalities:

1. Mental Health Prediction and Classification:
   • Utilizing a fine-tuned RoBERTa model trained on over 54,000 Reddit posts from SuicideWatch and Mental Health Collection datasets, WellMind classifies diary entries into categories like anxiety, depression, bipolar disorder, or suicidal thoughts. Predictions with a confidence level above 70% are analyzed to generate personalized feedback.
2. Personalized Advice:
   • Guided by evidence-based mental health resources from reputable organizations like the World Health Organization (WHO) and NHS, the platform uses Retrieval-Augmented Generation (RAG) to generate tailored advice for users. Semantic similarity comparisons ensure the advice is relevant, empathetic, and actionable.
3. Sentiment Analysis:
   • Diary entries not classified under specific mental health categories are analyzed for sentiment trends using VADER. Based on the sentiment—positive, neutral, or negative—the system offers supportive feedback or encourages users to elaborate further.
4. Interactive Dashboard:
   • Quotes Widget: Displays rotating motivational quotes every 10 seconds.
   • Streak Widgets: Tracks and displays consecutive diary entry streaks, motivating consistent engagement.
   • Wellness Tracking: Visualizes users' mental health trends over 30 days with a scoring system based on diary entry classifications.
   • Calendar Integration: Highlights diary entry days with celebratory emojis to encourage mindfulness habits.
5. Web-Based Platform with Scalable Architecture:
   • The frontend, built with React and Material-UI, offers a sleek and user-friendly experience.
   • The backend, developed using Python and FastAPI, ensures robust data handling with a microservices architecture.
   • Secure authentication using OAuth2 with JWT ensures user privacy and data security.
6. Admin Features for Organizations:
   • Admins can monitor user trends, reset passwords, delete accounts, and export anonymized user data in CSV format. This facilitates organizational use cases such as monitoring workplace wellness or supporting student mental health initiatives.
7. Diary Features:
   • Users can log, update, and delete diary entries while receiving real-time, AI-driven advice to enhance their mental health awareness.

Why did you create this project?

The global mental health crisis continues to grow, with millions struggling to access timely and effective support. Stigma, limited resources, and lack of personalized care often leave individuals without the tools they need to navigate their mental health journeys. Recognizing these barriers, the WellMind team sought to create a scalable, accessible solution that leverages cutting-edge AI to empower individuals and organizations to address mental health proactively.

WellMind bridges the gap between mental health care and technology by combining evidence-based insights with advanced machine learning techniques. The application promotes inclusivity, privacy, and reliability, ensuring that mental health support is universally available and tailored to each user’s unique needs. Whether helping individuals manage their day-to-day wellness or providing organizations with tools to monitor and support users, WellMind represents a transformative approach to tackling mental health challenges.

How does it work?

1. Data Collection and Preprocessing:
   • Data from Reddit's SuicideWatch and Mental Health Collection datasets were tokenized and padded using the RoBERTa tokenizer. Synonym augmentation with nlpaug addressed class imbalances, retaining a preprocessing setup that maximized model performance.
2. Model Training:
   • Multiple models (e.g., BERT, RoBERTa, DeBERTa, and MentalBERTa) were tested, with RoBERTa demonstrating the best classification accuracy. Fine-tuning with pre-trained embeddings allowed for robust predictions.
3. Advice Generation:
   • OpenAI's text-embedding-ada-002 model enables semantic comparisons between user inputs and a vector database of predefined suggestions. Advice generation thresholds ensure high relevance and coherence, and fallback mechanisms prevent potentially harmful suggestions.
4. Backend and Frontend Integration:
   • MongoDB stores user data securely, while AstraDB manages vector embeddings for advice generation. React provides an intuitive interface, and FastAPI facilitates seamless communication between components.

Impact and Vision

WellMind not only offers immediate mental health support but also contributes to societal change by increasing awareness and access to mental health resources. Its scalable design ensures the platform can serve diverse communities, from individual users to organizations like schools, workplaces, and healthcare providers. By making mental health care intuitive, accessible, and data-driven, WellMind aims to empower users worldwide to take control of their mental well-being.

This project underscores the potential of technology to address pressing societal challenges, fostering a future where mental health care is not only a priority but also a shared responsibility.