"Health Is the Real Wealth": Unsupervised Approach to Improve Explainability in Health-Based Recommendation Systems

Bartolomé Ortiz Viso, Carlos Fernandez-Basso, Jesica Gómez-Sánchez, María J. Martín-Bautista

Abstract: Nutritional recommendation systems are one of the major challenges in the field of recommendation systems. These systems can be based on various aspects such as individual preferences, group affiliations, or nutritional needs. The latter, known as healthy recommendation systems, aims to offer a menu tailored to the user and their vital needs, as well as having a positive impact on their health. However, although these systems allow for very precise multiple nutritional adjustments, users often do not understand what values are taken into account and why these values matter. This study proposes an unsupervised pipeline that generates nutrient-focused natural explanations, based on the nutritional data of the recommended recipes and nutritional textual guidelines made by experts.