Neural restaurant-aware dish recommendation

Food is the first necessity of the people. Due to the fast-paced modern life, people usually choose to dine out for convenience. While existing methods have paid efforts for the food recommendation, they are mainly limited in inferring users' personal preferences for online recipes, and ignore the dish ordering process in dine-out scenarios. Given the same recipe, different restaurants may produce various tastes due to food cuisines or chefs' cooking habits. In the current restaurant, users' general favored dish may have bad word-of-mouth. Thus, apart from their personal taste preferences, users also turn to restaurant specialties to guarantee the dish quality. As such, the restaurant-related dish quality and users' personal taste should be considered simultaneously. To address this task, we propose a neural restaurant-aware dish recommender to infer users' preferences for dishes in a specific restaurant. Given a dish in the current restaurant, whether to order it or not is mainly decided by two factors: users' personal taste and the dish quality in this restaurant. Our proposed model can: 1) capture users' personal diet preferences by the strong expressiveness of neural networks; 2) evaluate how good the current restaurant is at cooking certain dishes. To show the effectiveness of our proposed model, we conduct extensive experiments on a real dataset, demonstrating significant improvements over the several competing models, such as NCF with an average improvement of 36%, and PITF with 3.4%.