Azure Document Intelligence (Repository Option)

Grooper offers robust integration with Microsoft Azure Document Intelligence, enabling advanced cloud-based document analysis and optical character recognition (OCR) for a wide variety of document types. This integration streamlines the extraction of text, layout, and semantic data, supporting both automation and AI-driven workflows.

Overview of Azure Document Intelligence in Grooper

Azure Document Intelligence is a cloud service from Microsoft that provides intelligent document processing capabilities, including text extraction, layout analysis, and semantic understanding. Grooper connects to a Document Intelligence service by enabling and configuring the Azure Document Intelligence Repository Option. This is configured on the Grooper database Root node and provides connectivity by entering an API key and a resource name.

With the Azure Document Intelligence option added and configured, Grooper leverages the Document Intelligence service in two primary ways:

• The Azure DI OCR engine - Used for text extraction and layout data collection by the Recognize activity.
• The DI Analyze activity - Used for comprehensive document analysis that can be leveraged by Grooper's AI-enabled features (including AI Extract).
  • This analysis results in a JSON data file that is used by the DI Layout Quoting Method when configuring AI-enabled features.

Key similarities and differences between DI Analyze and Azure DI OCR

Similarities

Azure DI OCR and DI Analyze have several things in common.

• Both utilize the Document Intelligence service Grooper connects to using the Azure Document Intelligence option added to the Grooper Root.
• Both have access to the same models (although they utilize them differently).
  • Be aware, Grooper's current integration with Azure Document Intelligence has focused on using the prebuilt-read and prebuilt-layout models.
• Both can process page images and a Batch Folder's attachment file.

Differences

While both methods utilize Azure Document Intelligence, they differ in scope, output, and intended use:

Azure DI OCR

• Focuses on OCR (text recognition) for machine and hand print and layout data collection.
  • Using the prebuilt-read will perform text recognition only. Using the prebuilt-layout model layout data is also collected. Lines, checkboxes and (optionally) barcodes will be saved to the layout data file created by Recognize.
• Configured as an OCR Engine within Grooper's OCR Profile.
• Results can be used with Grooper's Value Extractions (Pattern Match, Labeled OMR, Labeled Value, etc.)
• Aligns Azure OCR results with Grooper's internal OCR engines for enhanced accuracy.

DI Analyze

• Performs full document analysis, extracting text, layout, style, and semantic data.
• Enables advanced AI workflows, including LLM prompt injection and "spatial grounding" to improve document highlighting when aligning an LLM's response back to the Grooper document.
• When run on the folder level, can be configured to prefer the folder's child pages (default) or attachment file.
• When run on the folder level, DI layout data is saved to both the folder and its child pages.
• Using the DI Layout Quoting Method, AI-enabled features can access results in text, markdown and HTML formats.
• Results cannot be used with Grooper's Value Extractions (Pattern Match, Labeled OMR, Labeled Value, etc.).