Published Paper Details:

In the rapidly evolving landscape of cultural tourism, offering real-time, multilingual, and personalized insights plays a pivotal role in enhancing the visitor experience, especially at historically rich and diverse locations such as those found across Tamil Nadu. This project presents an AI-powered voice assistant mobile application specifically designed to make heritage exploration more engaging, educational, and accessible for a wide range of users, including domestic and international tourists. The system integrates advanced natural language processing and machine learning technologies to interpret user queries in multiple Indian languages, whether spoken or typed, and respond with contextually relevant heritage information. At its core, the application employs a fine-tuned Mistral 7B transformer model to generate dynamic, human-like responses across various languages, while OpenAI's Whisper model ensures high-accuracy speech-to-text conversion for real-time voice interaction. Complementing this, the Google Text-to- Speech (gTTS) engine synthesizes natural-sounding audio responses, making the interface fully voice-driven and inclusive. Personalization is achieved through a GRU-based behavioral model that tracks and analyzes sequential user interactions, learning individual preferences to deliver tailored content and recommendations. The application incorporates real-time geolocation data and applies the Haversine distance formula to determine the user's proximity to various heritage sites, thereby offering hyperlocal suggestions and site-specific content. All user preferences, historical interactions, and metadata are efficiently managed using an SQLite database, ensuring smooth offline support and optimized performance in resource-constrained environments. Drawing from a structured heritage dataset in JSON format, the assistant provides rich narratives, historical facts, travel guidance, and site-specific tips, creating an immersive cultural experience.