Advanced Online Learning Options window
This window is used to customize the way online learning is deployed.
- General Settings
-
This group has the following options:
- Group by field
-
Select from the list the field value you want to use to tag the group to facilitate sorts and filtering. For example, you may use the name of a supplier as the group value. The value of this field is saved in an additional data field. The value for this option is set to <Classification Result> by default.
- Use Classification Online Learning
-
Select this option if you want to use online learning to improve your classification results over time. This is a project-wide setting that affects all classes in your project. This option is selected by default.
This is available only for projects created using the Classification Group template.
- Use Extraction Online Learning
-
Select this option if you want to use this type of online learning. This is a project-wide setting that affects any class with one or more trainable locator. This option is selected by default.
Important If this option is selected, the Use for Extraction Online Learning button is displayed in the Validation toolbar.This is available only for projects created using the Extraction Group or Shared templates.
- Classification Online Learning Settings
-
This is available only for projects created using the Classification Group template.
This group has the following options:
- Maximum documents stored for import
-
This option is used to determine the maximum number of documents that are stored for the Classification subset of the New Samples document set by the Knowledge Base Learning Serversystem task.
Enter a number between 100 and 20,000 to limit the number of documents to store for import. The value for this option is set to
2000
by default. - Use dynamic classifiers during classification
-
This option is used to create dynamic classifiers for your project. This option is selected by default.
If selected, all of the documents marked for Classification Online Learning are used by the dynamic classifiers when a document is classified. This means that any documents collected since the last time you trained your project do not have to wait until you train again to be useful.
- Extraction Online Learning Settings
-
This is available only for projects created using the Extraction Group or Shared templates.
This group has the following options:
- Maximum documents stored for import
-
Enter a number between 100 and 20,000 to limit the number of documents to store for import. The value for this option is set to
2000
by default. - Use dynamic knowledge base during extraction
-
If this option is enabled, a specific dynamic knowledge base is created by the system task for the documents marked for Extraction Online Learning, and this dynamic knowledge base is used next time extraction is performed.
This option is selected by default.
When disabled, the project administration has to import and review the documents returned from online learning and update the project before any improvements are made.
- Automatic training after Validation
-
Select to automatically flag a document for Extraction Online Learning. When the document is flagged it can be imported to improve the project. If the Use dynamic knowledge base during extraction option is selected a flagged document is added to the dynamic specific knowledge base and can be used during extraction in server processing. This option is cleared by default.
However, you have to define fields in the Field details window that are monitored for flagging by selecting the Monitor for automatic learning option. You can only monitor fields that are assigned to a trainable locator field. A document is used for Extraction Online Learning when the confidence of an extracted field is below a certain confidence level or field coordinates were changed during Validation or Thin Client Validation.
Definitions for the buttons at the bottom of this window can be found in Common Transformation Designer Buttons.