SharedRecognitionTools actions

The Datacap CCO is a file that contains a representation of recognized text. It contains all of the text that has been recognized along with additional meta data such as character positions and character confidence values. The confidence values indicate if the recognition engine is pretty sure that the recognized text is correct or a low confidence indicates that the result is a best guess.

The CCO is used by several different aspects of a Datacap application. A CCO is required when using a verification panel to allow a verify operator to identify text on a page using the functionality called click-n-key. One example of use is in the APT application where page data that was not found automatically by the application can be manually located and assigned by the verification operator by clicking on the target field and then clicking on the location of the text on the displayed image.

The CCO information is also used by some of the actions within a Datacap application. The Locate action library uses the CCO to find text on a page and associate the text with a field location. The found data can also be used to perform page identification or stored for later use by the application. One example could be finding the word "Date", or some form of the word, on a page and then take the text next to it and assigning it to a date field. Another example could be finding one or more keywords, such as "Invoice" and assigning a type to the page based on the keywords that have been found. Refer to the help for an action to determine if it requires a CCO to operate.

A CCO does not always contain text. For example, the action Analyze Image creates a CCO that is a representation of the image geometry. The use of this kind of CCO is to perform fingerprint matching. Fingerprint matching is a mechanism that performs page identification and also assigns a zone fingerprint template to the image The specific page type allows specific actions to be performed on the page based on the application rules while the fingerprint template identifies the location of the fields on the page. Fingerprinting can be performed via a CCO that was created by AnalyzeImage or via recognition.

A CCO is created through an action that can create a CCO. As described, there are several ways a CCO can be created. The actions RecognizePageOCR_A and RecognizePageOCR_S are examples of actions that directly produce a CCO file that contains the results of full page recognition.

Some actions perform full page recognition and store the results in an intermediary file called the "layout XML". This layout XML contains all of the recognized page data along with the character position information and confidence information. The layout XML file can be converted to a CCO using the action CreateCCOFromLayout. Once the CCO exists, then the recognition results can be used by subsequent actions as described previously.

Actions that can produce the layout XML include Convert.PDFFREDocumentToImage, OCR_SR.Recognize and RecognitionOCRA.Recognize. To use the results of these actions, call CreateCcoFromLayout on each page level DCO node that contains a layout file to create the CCO file.

Actions that directly create a CCO can filter out large text. The intent is to avoid having text from logos or large headings within the CCO which can change the alignment of the text. If text filtered out because, it is too large, it is logged in the action log file. The actions SetMaxCharacterHeightTMM and SetMaxCharacterHeightAVG allow adjustment of this process or can turn it off so large text is not filtered out.

The SharedRecognitionTools action library has other actions that can be useful to manipulate recognized text. The SnapDCOToCCO and SnapCCOToDCO actions can move data from the CCO into fields or data from fields into the CCO. SetFullPageRecognitionArea can adjust the area on the page that is recognized when full page recognition is performed. The OMR threshold actions can be used to determine if an area of the page contains an "X" or filled in box or oval, known as optical mark recognition.