Skip to main content

Amazon Textract

Amazon Textract

Get the insights you need. Extract, analyze, and understand text, handwriting, and data from virtually any document, with no model training required.

Why Amazon Textract?

Amazon Textract is a machine learning (ML) service that extracts text, handwriting, forms, tables, and layout from a wide range of documents. Amazon Textract extends beyond optical character recognition (OCR) to analyze and structure the contents of a page, returning the elements it identifies along with their location and a confidence score. Amazon Textract Insights introduces document understanding: describe the fields you need in natural language and receive structured insights from mixed, multi-document packets, with page-level grounding and a confidence score, without the need for model training.

Benefits

    Go beyond OCR to extract text, handwriting, forms, and tables with high accuracy. Amazon Textract Insights analyzes mixed, multi-document packets and returns the insights you describe in natural language, including values inferred from content and grounded answers you can verify.

    Add document understanding through managed APIs with no model training.

    Scale document processing up or down to match changing volume.

    Securely automate data processing with data privacy, encryption, and compliance standards.

Features

Abstract illustration of a translucent glass block casting a long shadow in front of a peach sun on a warm, rainbow-striped horizon.

Amazon Textract Core

  • Detect printed text in 40 languages, plus handwriting, and extract forms, tables, layout, signatures, and query answers, each with its location and a confidence score
  • Specialized APIs handle invoices, receipts, identity documents, and mortgage packages
Abstract illustration of iridescent blocks floating above a circular, multicolored checkered platform under a rainbow sky.

Amazon Textract Insights

  • Returns the insights you describe in natural language, including answers a document does not state outright, layouts, tables, charts, and multi-document packets up to 3,000 pages
  • Every answer returns a confidence score and page-level grounding for improved automation workflows and time-saving document processes
  • Get started fast with AWS blueprints or define your own

Use cases

    Process loan and mortgage applications more efficiently across financial documents such as a W-2 or Form 1040. Amazon Textract Core reads printed and handwritten text, captures form fields as linked key-value pairs, and preserves the rows and columns of tables. It extracts the data you need, such as applicant names, rates, and invoice totals, and returns each item with its location on the page and a confidence score, all without building a custom pipeline or training a model.

    Better serve patients and insurers by processing intake forms, insurance claims, and pre-authorization forms together. In a single request, Amazon Textract Insights splits a mixed submission into its separate documents and extracts the patient and claims data you describe in natural language, including answers inferred from the content. Each value carries a confidence score and page-level grounding, so you can set a confidence threshold, process the values that clear it, and route the rest for human review.

    Government agencies process a wide range of forms that arrive in many formats and layouts, from tax forms to small business loan applications and business filings. With Amazon Textract Insights, describe the fields you need in natural language and get insights back as structured data, without training a model. Start from AWS public blueprints such as Form W-9, Form 940, or Form 941. When a submission combines several documents, Textract Insights splits and matches them in a single call and extracts the data from each one. Each value comes back with a confidence score and page-level grounding to its source enabling more informed decision making.