Goal: You are a world-class expert in culinary recipe classification. Your task is to…
- User
Goal: You are a world-class expert in culinary recipe classification. Your task is to categorize a recipe accurately based on its title, description, ingredients, and instructions.
Context: A recipe can either be a beverage or a food item. If it's a beverage, classify it as "drinks". If it's food, choose the most appropriate category among the following:
"appetizer" – small dishes served before the starter
"starter" – light dishes served at the beginning of a meal
"main" – main course or substantial dishes
"dessert" – sweet dish served after the main meal
"breakfast" – dishes typically eaten in the morning
"snack" – quick, small bites or nibbles
"teatime" – food items commonly enjoyed with tea or coffee
"jam" – fruit preserves
"bread" – all types of bread
"sauce" – condiment or accompaniment to other dishes
"icecream" – frozen desserts based on cream or milkInput: You will receive structured data for a recipe in the following format, including title, description, ingredients list, and step-by-step instructions.
json {"title":"LENTILLES À LA DIJONNAISE null","description":null,"ingredients":[{"section":"For the recipe","ingredients":["800 g de lentilles cuites au naturel","200 g d'allumettes de jambon","3 oignons","40 g de beurre","1 cuil. à soupe de farine","1 cube de bouillon de bœuf","2 cuil. à soupe de moutarde forte","sel, poivre"]}],"instructions":["Épluchez et hachez les oignons.","Versez 50 cl d'eau bouillante sur le cube de bouillon.","Faites fondre le beurre dans une casserole, ajoutez les oignons et le jambon, saupoudrez de farine et délayez avec le bouillon.","Incorporez ensuite la moutarde, salez et poivrez.","Ajoutez les lentilles, mélangez bien, réchauffez et servez très chaud."]}
Your task:
- Analyze all fields (especially ingredients and instructions) for implicit clues about the type of dish.
- Avoid over-relying on the title alone.
- Choose the most precise and relevant category from the list.
- Always prefer specificity over generalization.
Output format: Return your answer strictly in the following JSON format:
json { "data": "" }
- Assistant
{ "data": "main" }