Recognize RecognizeRecognizes the internal content of a document or page, and saves the resulting data for use by subsequent activities.
This activity detects and reads the presentation elements which convey information in a document, such as text segments, barcodes, lines, check boxes, and other shapes. The resulting character data and layout information is saved on the Batch Folder or Batch Page object being processed, where it is available to subsequent activities which depend on recognition results.
Recognize should be executed in a Batch Process after any permanent image cleanup has been applied, and before any activities which depend on recognition output. (See further discussion below.) It can be executed at the page level or at the folder level.
When executed at the page level, recognizes one page at a time. If the input page has PDF content (i.e. was created by splitting a PDF document), then the page will be handled as a PDF page for the purposes of native text extraction. Otherwise, the page will be processed as an image, and text extraction will be purely OCR-based.
Page-level processing is the preferred operating mode in many cases, as it maximizes the benefits from parallel processing.
When executed at the folder level, recognizes all pages of a multipage document at once. The input folder must have a PDF or image-based document attached. To recognize other document formats such as Microsoft Word, HTML, and etc., use the Render activity to generate a PDF version prior to executing this activity.
Due to the CPU-intensive nature of recognition, folder-level processing may be unsuitable for large documents. For example, a single task running recognition on a 1000-page document could take 20 minutes to complete. This can result in long-running tasks which appear hung and services which are difficult to start and stop. In such cases, split the document into pages (see Content Action) and run recognition at the page level.
Any activity which accesses the internal content of a document will require recognition results in order to function properly. The following are specific examples of cases where other activities depend on the output from Recognize:
Inherits from: Unattended Activity
The following 9 properties are defined.
| Property Name | Description |
|---|---|
| General | |
| OCR Profile | Type: OCR Profile
Specifies the OCR Profile to be used for text recognition. The OCR Profile specified here is used to perform full-page OCR, and to detect layout objects such as lines, check boxes, barcodes, and shapes. The IP Profile assigned to the OCR Profile should include IP Commands which detect the set of layout objects needed for data extraction. See the 'Layout Detection Summary' property for more information. |
| Page Filter | Type: String
Restricts recognition to specific page numbers. If this value is blank, recognition will be performed on all pages. Otherwise, recognition will only be performed on pages specified in the filter. |
| Layout Detection Summary | Type: String
Summarizes the non-text features to be detected based on the configured OCR Profile. Layout Data is information describing the non-text elements of a page, such as lines, check boxes, barcodes, and shapes. It is important to capture and save layout data in cases where data extraction logic relies on the location of these elements. Layout Data is generated when the OCR Profile is configured with an IP Profile containing certain IP Commands. For example, Line Detection, Line Removal, Box Detection, Box Removal, Barcode Detection, Barcode Removal, Shape Detection, and Shape Removal are all examples of commands which generate layout data. Generally speaking, it is best to use the dropout flavor of these commands during pre-OCR image processing. |
| PDF Options | |
| Native Text Extraction | Type: TextExtractMode, Default: Full
Specifies how text should be extracted from PDF files. Can be one of the following values:
Note that the text extraction process operates against all text objects drawn on the page - whether they are actually visible or not. Unexpected results can occur when the input document contains text drawn transparently or behind other objects. This setting is only applicable if the input is PDF content. When running at the folder level, this means that the input document must be a native PDF document, or must have a PDF version generated by the Render activity. When running at the page level, this means that the page object must have been created by splitting a PDF document. |
| OCR Assist | Type: OcrAssistMode, Default: Auto
Specifies the extent to which OCR will be used in conjunction with Native Text Extraction. Can be one of the following values:
|
| Alternate IP | Type: IP Profile
An optional IP Profile used for detecting layout elements on PDF pages, in cases where OCR will not be used. The IP Profile specified here should contain IP Commands such as Line Detection, Box Detection, Barcode Detection, and Shape Detection. This IP Profile is used for detection purposes only, and the output image will be discarded. |
| Processing Options | |
| Error Disposition | Type: IssueDisposition, Default: Flag, Log
Determines what happens when an error occurs processing an activity. |
| Maximum Consecutive Errors | Type: Int32, Default: 0
The maximum number of consecutive errors, after which a critical stop will be raised. A critical stop will cause services to stop running. |
| Concurrency Mode | Type: ConcurrencyMode, Default: Multiple
Specifies the parallel processing mode for this activity. Can be one of the following values:
|