The correct choice is Amazon Textract. It extracts printed and handwritten text from scanned documents and forms and produces machine readable output that can be indexed and searched.

Amazon Textract is a managed OCR service that can detect handwriting and capture structured data such as form fields and tables and it eliminates the need to build and maintain custom OCR models when processing thousands of pages each month.

Amazon Comprehend analyzes the meaning and relationships within already digitized text and it does not perform OCR on images or PDFs so it cannot extract handwriting from scanned permits.

Amazon SageMaker lets you build and train custom machine learning models and you could create an OCR pipeline there but that requires extra development and maintenance compared with using a managed handwriting capable OCR.

Amazon Rekognition provides image and video analysis and basic text detection in images but it is not optimized for document oriented OCR or extracting structured handwritten fields from forms.

Full AWS Practitioner Certification Question