# Amazon Textract

> An AWS service that reads scanned documents and pulls out text, forms and tables.

- Level: Intermediate
- Topics: AWS, AI
- URL: https://tidbit.legios.com.ar/amazon-textract/
- In Spanish: [Amazon Textract](https://tidbit.legios.com.ar/es/amazon-textract/)

**Amazon Textract** takes an image or a PDF of a document, such as a scanned form, a receipt or a bank statement, and returns its contents as data a program can use. At the simplest level it does **OCR**, optical character recognition: it finds the typed and handwritten words and where they sit on the page.

What sets it apart from plain OCR is that it understands layout. It can read a **form** and pair each label with its value ("Name: Ana Pérez"), rebuild a **table** row by row and cell by cell, and answer **queries** like "what is the policy number?" even when the field is in a different spot on every document. Specialized operations handle invoices and receipts, and identity documents like U.S. driver's licenses and passports.

Typical uses are automating paperwork: loading invoices into accounting systems, processing insurance claims or loan applications, or making archives of scanned files searchable. Each result comes with a confidence score, so a common pattern is to let high-confidence fields through automatically and send the doubtful ones to a person to review.

## An analogy

A data-entry clerk who takes a stack of paper forms and types each one into a spreadsheet, putting every value in the right column, and puts a sticky note on any field they could not quite read.

## Keep in mind

- Extracts typed and handwritten text from images and PDFs.
- Understands forms (label and value) and tables, not just loose words.
- Queries let you ask for a specific field by question.
- Confidence scores help decide what still needs human review.

## Go deeper

- [What is Amazon Textract? (official docs)](https://docs.aws.amazon.com/textract/latest/dg/what-is.html)
- [Amazon Textract product page](https://aws.amazon.com/textract/)
- [Optical character recognition on Wikipedia](https://en.wikipedia.org/wiki/Optical_character_recognition)

## Related terms

- [OCR](https://tidbit.legios.com.ar/ocr/): Software that reads the text in an image so a computer can search and edit it.
- [Spreadsheet vs database](https://tidbit.legios.com.ar/spreadsheet-vs-database/): Why a company that runs on Excel eventually needs a real database.