Classification tab - Project Settings window

Closes the window and saves your changes.

Closes the window without saving your changes.

Displays the help for the open window.

Default classification result

Select the desired default classification result that is used if no other classification result can be determined. (Default: unclassified)

Classification evaluation

Select the type of classification evaluation you want to use. The following choices are available:

• Automatic evaluation (Default: Selected): The specified values for confidence and distance are used to evaluate the result. For multi-page documents the classification is performed page by page and stops when a page can be classified. All pages following the classified page are not processed.

• Script implemented evaluation: The same page by page classification loop is executed, but your custom script is responsible for evaluating the classification results, breaking the classification loop, and determining the final classification result for the document. You cannot use class-level classification along with script implemented evaluation. If you select the script implemented evaluation value, any previously-made settings are ignored.

Classification behavior

Choose how a document is classified using layout classification from one of the following choices:

• Classify first page only (Default: Selected): Choose this value if you want the classification loop to processes only the first page of a document for the specified classifier type. If the page cannot be classified the document remains unclassified.

• Classify each page: Choose this value if you want the classification loop to processes each page of a multi-page document for the specified classifier type until a page can be classified or the complete document is processed and remains unclassified.

• Do not use layout classification: Choose this value if you do not want to use the Layout Classifier.

Minimum confidence

Use this setting to set the minimum confidence. This is the smallest value required for automatic evaluation to assign a classification result. (Default: 70)

Minimum distance

Use this setting to set the minimum distance between the best and the second best classification result. (Default: 10)

Layout Classifier

If layout classification is enabled for one or more classes in your project, this classifier is run during production. The following buttons are available for managing your Layout Classifier.

Button	Description
Properties	Displays the Layout Classifier Properties window.
Reset	Removes the Layout Classifier from your project. If layout classification is enabled for any class, this is skipped and content classification is performed instead.

Training behavior

You can ignore one or more pages of a document if you want to exclude them from training. This is done via the Document Viewer Thumbnails view.

Choose how a document is trained from one of the following values.

• Train first page only (Default: Selected): Choose this value if you want to train a document for content classification based on the first page that is not ignored.

  When this value is enabled, recognition for content classification is performed for the first page of a document only. If you later switch to "Train all pages" the project needs trained again. This ensures that recognition is performed for all pages so information is available for content classification and that text editing is possible.

• Train all pages: Choose this value is you want to include the complete document for content classification, excluding any ignored pages.

The selected training behavior is also used for Classification Online Learning.

Classification behavior

Determine how a document is classified using content classification from one of the following values.

• Classify first page only: (Default: Selected) Choose this value if you want the classification loop to process only the first page of a document for the specified classifier type. If the page cannot be classified the document is unclassified.

• Classify each page: Choose this value if you want the classification loop to processes each page of a multi-page document for the specified classifier type until a page can be classified or the complete document is processed and stays unclassified.

• Classify all pages at once: Choose this value to merge the text of all pages so it can be processed by the content classifier at once.

• Do not use content classification: Choose this value if you do not want to use the content classifier. This value can be used to avoid OCR on demand during classification if only the Layout Classifier is used.

After selecting the content classifier method, you can specify confidence with the following content classifier settings:

Minimum confidence:

Use this setting to set the minimum confidence. This is the smallest value required for automatic evaluation to assign a classification result. (Default: 50)

Minimum distance:

Use this setting to set the minimum distance between the best and the second best classification result. (Default: 10)

Content Classifier:

If content classification is enabled for one or more classes in your project, this classifier is run during production. You can select between the following content classifier types:

• Adaptive Feature Classifier (Default: Selected): This is a learn-by-example classifier that can be trained and improve classification results over time. The following buttons are available for this classifier

• Text Classifier: This is a deprecated classifier and should only be used for old projects that use this type of classifier already.

Use the following buttons to manage the selected content classifier:

Button	Description
Properties	Displays the Adaptive Feature Classifier Properties window for the main content classifier.
Reset	Removes the main content classifier from your project. If content classification is enabled for any class, the classification result relies on other methods of classification.

Classification behavior

Select one of the following values:

• Number of pages for classification: To use Generative AI classification, select this value and then specify how many pages are considered in the classification. (Default: 2)

  The different classification methods, if enabled, are run in the following sequence: Layout classification, content classification, and then Generative AI classification. Later methods are run only if previous methods do not produce a confident classification result. This means that Generative AI classification is attempted only if no classification result is available from the other enabled classification methods.

  To use this type of classification at the class level, the corresponding Generative AI classification setting in the Class details must be enabled for one or more classes in your project.

• Do not use Generative AI classification: If selected, no generative AI classification is performed. Instead, layout and content classification is used, if enabled. (Default: Selected)