Graph Neural Networks for Culinary Pairing: Enhancing Flavor Harmony in Food Recommendations

Abstract

The practice of blending ingredients to create harmonious flavors—known as culinary pairing—has long piqued the interest of the culinary community. A growing number of people are interested in using machine learning models to improve culinary experiences as a result of the development of computational tools in food science in recent years. In this work, we suggest applying Graph Neural Networks (GNNs) to the problem of flavor harmony in meal suggestions. GNNs allow us to capture complex linkages and interactions among ingredients by visualizing components and their flavor profiles as nodes and edges in a graph structure. We provide a novel method that combines sophisticated flavor embedding techniques with GNNs to produce personalized, culturally-appropriate food recommendations. The experimental results of the authors demonstrate the effectiveness of the proposed method in enhancing flavor harmony and providing tailored culinary suggestions, thereby enriching the gastronomic experiences of individuals.

DOI - https://doi.org/10.65525/SVUP.9788199778009.2026.34-43