> Relying only on ingredient and food names extracted from a recipe database would limit the ability to capture a broader range of culinary terms, since there could exist food-related assertions that do not necessarily contain those terms, which in result may narrow down the scope of the resulting assertions.
> To mitigate this, we also add general terms such as “Food” and “Nutrition” as well as specific terms such as “Vitamin B” for nutrition and “Diabetes” for healthy diet. We develop a two-level ontology, allowing us to categorize 1,600 culinary terms.
>We use ChatGPT3 to collect those common culinary terms. The prompt “Please provide an exhaustive list of verbs related to cooking actions and techniques, such as chopping, slicing, seasoning, and garnishing.” is used to collect common culinary terms. Additionally, to acquire the list of professional or scientific terms related to food, we utilize the prompt “I want to develop a dataset based on food computing and want to aggregate abstracts of related papers. Please suggest me 20 keywords that will provide such insights.”