//
Busque em todas as páginas da documentação
7 pages in this section.
Beyond OCR: forms, tables, and queries as first-class outputs, and the flat Blocks graph every Textract operation returns.
Extract raw text from a scanned document via SDK - your first DetectDocumentText calls in both Python (boto3) and TypeScript (AWS SDK v3).
Structured key-value and table extraction from documents via SDK - walking KEY_VALUE_SET and TABLE/CELL blocks with AnalyzeDocument.
Routing low-confidence extractions to human reviewers via SDK - HumanLoopConfig on AnalyzeDocument and the A2I flow definition it depends on.
A checklist for confidence-threshold handling, document-quality preprocessing, sync/async request patterns, and reliability when calling Textract from the SDK - across boto3 and AWS SDK v3.