A Multilingual AI-Powered Mental Health Chatbot Using Gemini API with Conversational Memory and Voice Interaction

Abstract

Mental health challenges have become increasingly common among students, working professionals, and individuals facing academic pressure, workplace stress, emotional instability, and social isolation [1], [2]. Due to social stigma, limited access to professional counseling, high consultation costs, and language barriers, many people hesitate to seek timely psychological support [3]. To overcome this problem an AI-powered multilingual mental health chatbot designed to provide immediate emotional assistance through natural and supportive conversation. The paper proposes using Google Gemini API integrated with a Flask-based backend and a responsive web interface, enabling intelligent human-like interactions in English, Kannada, and Hindi. To improve accessibility and user engagement, the chatbot incorporates voice input functionality using browser-based speech recognition, allowing users to communicate through speech in addition to text [4]. The system also maintains conversation history to preserve context and continuity. Unlike traditional rule based chatbots, the proposed model leverages generative artificial intelligence to deliver context-aware, empathetic, and language consistent responses without providing medical diagnosis [5], [6]. Experimental evaluation based on simulated user interactions demonstrates improved response relevance, multilingual consistency, and user satisfaction, with strong performance across standard evaluation metrics such as accuracy, precision, recall, and F1-score [7].