Data Normalization

Data processing and normalization tasks evaluate structured output from messy prose and different structures. This capability is essential for catalogue and product pipelines where data needs to be extracted from unstructured text and formatted consistently. Models must understand the desired output format and extract relevant information accurately.

Input Tokens

Output Tokens

Duration

Cost

Context

Input

Analyze the following cooking instructions and return a structured response following these criteria:
1. Identify all ingredients used in the recipe, even if optional
2. Classify each ingredient into one of three categories: 'main', 'side', or 'sauce'
3. 'Main' includes core components of the recipe; 'side' includes extras served alongside or supporting ingredients in the preparation; 'sauce' includes only sauces not condiments like salt
4. Use generic names when appropriate (e.g., any bread = 'bread')

Expected output

{
  "main": [
    "mince",
    "bacon",
    "cheese",
    "bread",
    "lettuce",
    "tomato",
    "onion"
  ],
  "side": [
    "potatoes",
    "oil",
    "salt",
    "pepper"
  ],
  "sauce": [
    "mustard",
    "ketchup",
    "mayonnaise"
  ]
}

Model output

{
  "main": [
    "mince",
    "burger patties",
    "cheese",
    "bread",
    "bacon",
    "lettuce",
    "tomato",
    "onion"
  ],
  "side": [
    "potatoes",
    "fries",
    "oil",
    "salt"
  ],
  "sauce": [
    "mustard",
    "ketchup",
    "mayonnaise"
  ]
}