Skip to content

Amazon OCR

Overview

The Amazon OCR Service in AIForged uses Amazon Textract to extract raw text from documents, without preserving layout information. The extracted text is stored in the Result property of the document for downstream processing, analysis, or integration.

Info

Use this service when you need plain text extraction from images or PDFs for workflows such as data mining, regex pattern detection, and downstream AI processing.

Supported Content Types

  • PDF
  • JPEG
  • PNG

Info

If your content is in a different format, use the AIForged PDF Converter to generate a compatible file.

Possible Use Cases

  • Extract text from scans of receipts, invoices, or contracts.
  • Run RegEx patterns on extracted text to detect keywords or structured data.
  • Prepare training datasets for machine learning by extracting plain text.
  • Index documents for search functionality.

Service Setup

Follow these steps to add and configure the Amazon OCR Service to your agent:

  1. Open the Agent View Navigate to the agent where you want to add the service.
  2. Add the Amazon OCR Service Click the Add Service button.
  3. Select Service Type Choose Amazon OCR Service from the available service types.
  4. Configure the Service Wizard Open the Service Configuration Wizard. or
  5. Step 1: General Settings Configure core OCR settings such as service name and description. &#xNAN;Default settings are sufficient for most use cases.
  6. Step 2: Verification No verification is typically required for text extraction services unless paired with downstream validation.

Service Configuration Settings

Most users can proceed with default settings; however, advanced settings can be adjusted for custom requirements:

Setting Type Required? Description
ArchivingStrategy Optional No Days before documents are deleted.
AccessKey Optional No Override the AWS Access Key.
BaseURL Optional No Override the AWS endpoint URL.
BatchSize Hidden - Size of processing batches.
CheckElectronic Optional No Detect if a document is electronic and extract text accordingly.
DocumentProcessedStatus Optional No Status applied when processing is complete.
Enabled Hidden - Enables or disables the service.
ExecuteBeforeProcess Optional No If a child service, run before the parent.
ExecuteAfterProcess Optional No If a child service, run after the parent.
IsVerification Optional No Save OCR results for human verification.
Language Optional No Specify a language hint for OCR.
MinSize Optional No Minimum document size to trigger OCR.
Password Optional No Authentication password (can be set per document).
RemoveComments Optional No Remove human comments before processing.

Info

If unsure, keep defaults unless you have a specific processing or integration requirement.

Add and Process Documents

To upload and process documents using the Amazon OCR Service:

  1. Open Service In the Amazon OCR Service, click the button or drag and drop files over the document grid.
  2. Select Category (Optional) If you know the category for the document, select it. Otherwise, select No category.
  3. Process Documents After uploading, select the documents to process and click Process Checked.

Info

Tip: For new services, process a small batch first to confirm OCR accuracy before scaling up.

View Processed Documents

  • In the Amazon OCR Service, set the usage filter to Outbox.
  • Open any processed document to view the extracted text in the Result property.

Troubleshooting Tips

  • Text Missing or Incorrect?
  • Check document quality and resolution (300 DPI is recommended).
  • Specify a language hint in the configuration to improve recognition.
  • Ensure the correct file type is being used.
  • Slow Performance?
  • Large documents or batches may take longer — process smaller sets where possible.
  • Unexpected Characters or Symbols?
  • This can occur if OCR is trying to interpret complex fonts or handwriting.
  • Pre-process your documents (cleaning, adjusting contrast) before uploading.

Best Practices

  • Use high-quality scans — blurry or low-resolution images can reduce OCR accuracy.
  • Specify the language in the settings if most documents are in a language other than English.
  • Avoid uploading password-protected PDFs unless you supply the password in the configuration.
  • Remove non-essential marks or annotations that may cause OCR noise.

Known Limitations

The following provider constraints apply to Amazon Textract-backed OCR:

  • Supported formats
  • JPEG, PNG, PDF, and TIFF. XFA-based PDFs are not supported.
  • Size and page limits
  • Synchronous operations: up to 10 MB in memory; PDF/TIFF limited to 1 page.
  • Asynchronous operations: PDF/TIFF up to 500 MB and up to 3,000 pages.
  • PDF specifics
  • PDFs cannot be password protected
    • Use the AIForged Custom Code utility as a pre-processor to set the password per document, so that AIForged can unlock documents before processing.
  • Max page size: up to 40 inches (2880 points).
  • JPEG 2000 images embedded in PDFs are supported.
  • Image constraints
  • Max resolution supported is 10,000 pixels on any side.
  • Minimum detectable character height is \~15 pixels (approx. 8 pt at 150 DPI).
  • All in-plane rotations are supported.
  • Language and script
  • Printed text detection supports multiple Latin-based languages.
  • Handwriting recognition is supported for English only.
  • Vertical text layouts (e.g., vertical East Asian scripts) are not supported.
  • Language is not returned in the output for standard text detection.
  • Query features (if used)
  • Per-page limits apply (e.g., max queries per page in sync/async modes).

Info

Tip: For very large PDFs or TIFFs, consider chunking by page range to avoid hitting page or size limits and to keep processing responsive. Use the AIForged Document Splitter for this task.

Permissions Required

Members must belong to one of the following AIForged user group roles to add and configure this service:

  • Owner
  • Administrator
  • Developer

Info

Tip: Role membership is managed in Organisations > Roles. Assign members to roles to grant agent and service administration access.