Paper Title
Restaurant Menu Dish Recommendation System Using Vector Databases and Approximate Nearest Neighbor Algorithms

Abstract
- tAIsty is an intelligent restaurant menu recommendation system that combines content-based and collaborative filtering to help users discover dishes according to their preferences. It leverages Approximate Nearest Neighbors (ANN) techniques, like PyNNDescent, to efficiently identify similar dishes based on attributes such as cuisine, category, and ingredients. For collaborative filtering, the system tracks user interactions (clicks, orders, and feedback) and uses similarity-based matching to recommend dishes favored by users with comparable tastes. To support real-time and scalable recommendations, tAIsty integrates a vector database (MongoDB Atlas, Weviate) and serves content via FastAPI, with a modern UI in React. The system also includes thoughtful personalization features like allergy-aware filtering, ensuring safe and tailored suggestions. Furthermore, it incorporates an ingredient expiry tracking system to inform restaurants about upcoming shelf-life concerns, promoting dishes that reduce potential food waste without compromising quality. Designed for easy integration into digital menus, tAIsty aims to enhance the dining experience while supporting operational efficiency and sustainability. Keywords - Restaurant Recommendation System, Hybrid Filtering, Vector Search, Approximate Nearest Neighbors, PyNNDescent, Food Waste Reduction, Allergy Filtering, FastAPI