Data Extraction Pipeline

# Data Extraction Pipeline

You are a data engineering specialist. Design a prompt-based pipeline to extract structured data from unstructured text.

## The Extraction Task

Source document type: [e.g., PDF invoices, customer emails, news articles, job postings, product reviews]
What I need to extract: [e.g., company name, date, dollar amount, sentiment, named entities]
Volume: [DOCUMENTS PER BATCH]
Output format needed: [e.g., CSV, JSON, database rows]
Downstream use: [e.g., feed into CRM, analyze in spreadsheet, trigger another automation]

## Pipeline Design

**Step 1: Pre-processing** — How to clean and normalize the input before sending to the LLM.

**Step 2: Extraction prompt** — Write the exact prompt to extract the specified fields. Include the output schema as a JSON example.

**Step 3: Validation** — Rules to check that extracted values are plausible (type checks, range checks, required fields).

**Step 4: Error handling** — What to do with documents where extraction fails or confidence is low.

**Step 5: Output formatting** — How to convert the extracted JSON into the target format.

## Accuracy Tips

List 3 prompt engineering techniques to improve extraction accuracy for my document type.