Document classification script
An ABBYY FlexiCapture classifier processes submitted documents and determines their class. This lets you choose a document definition to be used for field extraction.
Each classifier corresponds to a specific classification training batch.
If you want to use several classifiers for a single batch, you need to set up a classification script. This can be useful when there are documents that need to undergo additional classification. In this case, the first classifier will be used to classify the documents according to their class, and then, an additional classifier will be selected from the script depending on the classification results. For example, documents can be first classified into "invoices" and "contracts", then the invoices can be further classified according to company name.
Note: This script can only be set up for an already existing classification training batch and cannot be set up when creating a new batch.
To set up a classification script:
- Open Project → Project Properties....
- Go to the Recognition tab found in the project properties window.
- Open the classification script Editor by clicking Set... to the right of Use script in the Classification section.
- In the dialog box, select an appropriate classifier by clicking Add... and choosing a classification batch from the list.
- Now, you can change a classifier's Alias Name. This allows you to not have to rewrite a script if a classifier's name has been changed during processing. The script will refer to the particular name specified in the Alias Name column.
- Open the script editor window by clicking Edit Script....
|A classifier in the batch type.
Classification confidence. Unlike working with automatic classifiers, when working with the script this parameter must be manually specified in page properties.
|The object for logging the information about processing.
Sample script that combines multiple classifiers for cascade classification
Sample script that assigns the Unknown class to documents, whose classification confidence turns out to be lower than the specified value