Optimizing Ingredient Substitution Using Large Language Models to Enhance Phytochemical Content in Recipes

Published: 16 Sep 2024

Authors: Luis Rita, Josh Southern, Ivan Laponogov, Kyle Higgins, Kirill Veselkov

Affiliations: Imperial College London; Boston Children’s Hospital; Harvard Medical School; Yale University

Introduction

In the emerging field of computational gastronomy, integrating cooking practices with scientifically supported nutritional goals is becoming increasingly important. This study explores how large language models (LLMs) can be applied to optimize ingredient substitutions in recipes, specifically to enhance the phytochemical content of meals.

Phytochemicals and Their Potential Health Benefits

Phytochemicals are bioactive compounds found in plants that, based on preclinical studies, may provide potential health benefits. These compounds have been shown to have positive effects on human health, making them an essential part of a healthy diet.

Methodology

We fine-tuned models including OpenAI’s GPT-3.5, DaVinci, and Meta’s TinyLlama using an ingredient substitution dataset. The models were used to predict substitutions that enhance phytochemical content and create a corresponding dataset of phytochemical-enriched recipes.

Results

Our approach improved the Hit@1 accuracy on the ingredient substitution task from a baseline of 34.53±0.10% to 38.03±0.28% on the original GISMo dataset, and from 40.24±0.36% to 54.46±0.29% on a refined version of the same dataset. These substitutions led to the creation of 1,951 phytochemical-enriched ingredient pairs and 1,639 unique recipes.

Conclusion

Although this approach shows potential in optimizing ingredient substitutions, caution must be exercised when drawing conclusions about health benefits, as these claims are based on preclinical evidence. Future work should include clinical validation and broader datasets to further evaluate the nutritional impact of these substitutions. This study represents a step toward using artificial intelligence to promote healthier dietary practices, providing a potential avenue for integrating computational methods with nutritional science.