A survey of food recommenders

TL;DR

This survey reviews the landscape of food recommender systems, covering their goals, data sources, technical methods, and constraints, with a focus on applications in health, wellness, and diverse dietary needs.

Contribution

It provides a comprehensive overview of food recommender systems, categorizing their types, datasets, algorithms, and system constraints, highlighting recent research trends and challenges.

Findings

Diverse datasets and signals are used for training food recommenders.

Various technical approaches include collaborative filtering, content-based, and hybrid models.

Food recommenders are increasingly tailored for health and dietary restrictions.

Abstract

Everyone eats. However, people do not always know what to eat. They need a little help and inspiration. Consequently, a number of apps, services, and programs have developed recommenders around food. These cover food, meal, recipe, and restaurant recommendations, which are the most common use cases, but also other areas such as substitute ingredients, menus, and diets. The latter is especially important in the area of health and wellness where users have more specific dietary needs and goals. In this survey, we review the food recommender literature. We cover the types of systems in terms of their goals and what they are recommending, the datasets and signals that they use to train models, the technical approaches and model types used, as well as some of the system constraints.