This article demonstrates how to automatically extract text from scanned PDFs and translate them while preserving original formatting using Amazon Textract, Amazon Translate, and Apache PDFBox.

• Amazon Textract extracts text and geometry data from scanned PDF documents
• Amazon Translate provides neural machine translation across 2,970+ language pairs
• Solution uses geometry-based approach to overlay translated text maintaining original layout
• Supports English, Spanish, Italian, Portuguese, French, and German extraction
• Open-source PDF Translate library available on AWS Samples GitHub
• Font size automatically calculated to fit translated text within original bounding boxes
• Processing time: ~10 seconds for employment application, under 1 minute for text-heavy documents
• Pay-as-you-go pricing based on pages processed and characters translated
• Optional human review workflows available via Amazon A2I for validation

The solution enables scalable, cost-effective multilingual document translation while maintaining document structure and formatting for regulatory compliance and human review.