Make a Table Locator trainable

A Table Locator that uses automatic extraction can be defined as a trainable locator. Select "Enable training" from the Settings tab to enable this functionality and train the table layout. Similar to other trainable locators a training set is created using the Edit Document window. You can either train against a document set or documents that are returned from Validation selected from the New Samples file list.

To reuse a training set for other projects you can create a knowledge base (*.kbtbl) for the specific training type from the Knowledge Base tab.

You can make a Table Locator trainable by following these steps:

  1. Open the locator properties.
  2. On the Settings tab, for the Detection mode, select Automatic.
  3. Select Enable training.
  4. On the Validation tab in the Project Settings, verify that Automatically extract empty tables in is cleared.

    If the Validation user manually indexes tables, and document table rows are automatically populated, the information might disrupt training.

  5. Load a training document from a custom test set or the New Samples document set.
  6. To add a selected document to the Extraction training set using the Edit Document window, press F10.
  7. Click Configure Table and define the table layout as follows.
    1. Draw the master item within the Document Viewer, to define the line item template.

      If needed, set additional options on the Master Item tab.

    2. Define the table columns on the Cells tab and draw an anchor, if needed.
    3. Click Test to test the configuration.

      The Test Results Tab displays the extracted rows.

  8. Click Validate to check the document for any problems.
  9. Add the document to the Extraction Set.

    The project is trained automatically.

  10. Open a test document and click Test to test your settings.
  11. Optionally, click Close to close the locator properties window.
  12. Save the changes to your project.