Planning Support System for Genre-specific Culinary Walking Tours Based on Dish-level Recommendations

This study introduces a planning support system for culinary walking tours focused on specific cuisine genres such as curry or ramen. Unlike conventional gourmet platforms where users explore menus after selecting a restaurant, the proposed system centers the planning process on individual dishes, recommending the next dish based on flavor similarity, diversity, and walking distance. It also enhances accessibility by automatically extracting dish names, descriptions, and images from unstructured review texts and images, allowing users to consider dishes even without structured menus. This dish-first, multi-restaurant approach reflects the exploratory and comparative nature of culinary walking tours. In a controlled user study, the proposed system tended to receive higher subjective ratings than Tabelog, which may indicate possible advantages in perceived diversity and enjoyment in this exploratory setting. This work contributes to the field of personalized food recommendation by modeling dish-level selection behavior and leveraging unstructured data to support distance-aware decision support during user-driven culinary exploration.