Abstract: A restaurant service system includes a restaurant server adapted to communicate with diner communication devices and restaurant communication devices. The restaurant server builds a food taxonomy including dishes, dish attributes and dish ingredients. The dishes are mapped from menu items. The server also trains dish classifiers and applies the dish classifiers to map menu items to dishes. Dish attributes are derived from search logs and dish ingredients are extracted from menu item descriptions. The server determines the search intent of a diner based on her search terms. Food search is conducted based on the taxonomy and the search intent with consideration of restaurant similarity and dish popularity. The server also provides dish recommendations based on taste similarity between diners.