Amazon Textract
- What it is
- Automatically extracts text, handwriting, forms, and tables from scanned documents and PDFs.
- Goes beyond OCR: it preserves the structure — key-value pairs and table cells, not just a string of characters.
- Specialized APIs
- AnalyzeExpense — invoices and receipts. AnalyzeID — driver’s licenses and passports.
- Queries — ask a natural-language question (“what is the policy number?”) and get the answer from the document.
- Common pipeline
- Amazon S3 → Textract (extract) → Amazon Comprehend (understand and redact PII) → database or search index.
- Not to be confused with
- Amazon Rekognition — general computer vision on photos and video; its text detection has no notion of forms or tables.
- Exam signal
- “Process invoices / forms / scanned documents” → Textract.