Amazon Textract
OCR++ — extract text, forms, and tables from documents.
❓ What is it?
A document-analysis service that goes beyond plain OCR: it preserves structure, returning key-value pairs from forms, rows and cells from tables, and purpose-built extractions for invoices, receipts, IDs, and mortgage documents.
💡 Why does it exist?
Businesses run on scanned paper and PDFs. Plain OCR gives you a wall of characters; Textract gives you STRUCTURED data — which field is the invoice total, which cell is quantity — making automation actually possible.
⏱️ When should you use it?
Use it to digitise document workflows: claims intake, invoice processing, KYC document reading. Pair with A2I when low-confidence fields need human verification before entering downstream systems.
🗺️ Where does it fit?
At the front of intelligent document processing (IDP) pipelines: documents land in S3, Textract extracts structure, Comprehend classifies or redacts, and results feed databases or workflow engines.
🔌 How do you integrate it?
Call AnalyzeDocument (forms/tables), AnalyzeExpense (invoices), or AnalyzeID synchronously for single pages; use asynchronous Start/Get jobs for multi-page PDFs in S3.
🧩 Commonly integrated with
🎯 Exam angle (AIF-C01)
- "Extract data from forms/tables/invoices" → Textract; plain "analyse the meaning of text" → Comprehend.
- Asynchronous APIs handle multi-page PDFs; synchronous calls are for single-page images — a recurring detail.