The goal is to categorize foods in a way that makes sense to users of the Food and Nutrient Database for Dietary Studies (FNDDS). Users include food retailers, nutrition/food science researchers, and consumers. Accurate categories can help people find foods and ingredients that meet their taste and nutrient needs. NLP categorization can be compared to clustering based on nutrient content. (Asking an existing LLM to do it would be costly, as there are hundreds of thousands of foods.)