Azure DI OCR

Recognizes machine print and hand print using the Azure Document Intelligence service.

Remarks

This OCR engine uses Azure Document Intelligence in combination with Grooper's internal OCR engines to generate detailed and highly accurate OCR results. It leverages the high accuracy of AI-powered recognition without sacrificing the detailed character-level positional data captured by traditional OCR engines.

Usage

Azure DI OCR is typically used in conjunction with the DI Analyze activity. In a batch processing workflow, the DI Analyze activity is executed first - either at the document or page level. This performs the Azure analysis and saves the results to the document or page as a JSON file.

For most use cases, the model used for analysis should be prebuilt-layout. This provides the word recogniton
necessary for Azure DI OCR to generate character-level OCR results, as well as the paragraph and table data needed for injecting HTML / markdown into data extraction operations. (See the DI Layout quoting method for more information on injecting layout data.)

, which includes paragraph, tables, figures and other layout information. This model provides the necessary data for Azure DI OCR to perform character-level recognition during the subsequent Recognize activity.

After DI Analyze, a Recognize activity runs the Azure DI OCR engine, which uses the stored Azure DI analysis data combined with traditional OCR engine results to generate character-level OCR data.

Standalone Use as OCR Engine

Azure DI OCR can be used as a stand-alone OCR engine. If no existing Azure data is found on the page or document, the engine will send the image to Azure and retrieve OCR results in real time.

In this mode, use the prebuilt-read model, as this model represents the low-cost "OCR-only" option.

Properties

Name	Type		Description
Model Name	String	►	The available document processing models. EditorModel Editor Requiredtrue Defaultprebuilt-layout Use the prebuilt-layout model if using the DI Analyze activity in the same process and injecting layout data into extraction operations using the DI Layout quoting method. Use the prebuilt-read model if using Azure DI OCR as a stand-alone OCR engine without a prior DI Analyze step.
Language	String	►	Specifies the language in which OCR will be performed. ConverterLanguage Converter DefaultEnglish Can be one of the following types: Value Description af sq ar ast az eu be brx bs br bg ca zh-Hans zh-Hant kw co hr cs da prs nl en et fo fil fi fr fur gl de kl haw hi hu is smn id ia iu ga it ja jv kea quc ko ku-Arab ky lkt lt dsb smj lb ms mt gv mi mr mn ne no oc os ps fa pl pt pa ksh ro rm ru sa gd sr sr-Latn sms sk sl so sma es sw sv tg tt to tr tk hsb ur ug uz-Arab uz-Cyrl uz wae cy fy zu If no value is provided, multi-language OCR will be enabled. Please note that Azure does not detect the language of documents. Run the Detect Language activity after extraction to ensure the proper culture code is assigned to the document.
Features	AzureFeatures	►	Specifies which Azure features to enable during OCR analysis. ConverterPg Flags Converter Defaultnone A combination of the following flags: Name Value Description none 0 No features enabled. barcodes 1 Extracts barcode values. formulas 2 Extracts mathematical expressions. keyValuePairs 4 Extracts key value pairs (form fields). languages 8 Enables language detection. ocrHighResolution 16 Performs OCR at a higher resolution to handle documents with fine print. queryFields 32 Enables the extraction of additional fields via the queryFields query parameter. styleFont 64 Enables the recognition of various font styles. Choose additional Azure features to include in the OCR process, such as barcode extraction, key-value pair recognition, language detection, and font style analysis. Enabling more features may increase processing time and result detail. See the AzureFeatures enumeration for available options.
Character Data Alignment	OCR Data Aligner	►	Configures alignment and correction of character data between Azure OCR results and Grooper's OCR engines. EditorObject Properties Editor ConverterExpandable Converter Use this property to specify how character data from Azure OCR is aligned and corrected using Grooper's internal OCR engines. Adjust alignment settings to improve accuracy for specific document types or layouts. For more information, see OCRDataAligner.
File Name	String	►	The name of the file where the Azure results will be saved on the page. The results of the Azure analysis are saved to each page as a JSON file. The file name is generated based on the selected model, following the pattern AzureDI_{ModelName}.json. This allows you to easily locate and review the output for each page.

See Also

OCR Data Aligner

Used By

OCR Profile