Train the Trainable Evaluator

By training this evaluator, you are teaching your project to extract the best results from input locators, even if those locators are not trainable themselves. For example, you can use a Format Locator to extract a customer number on a document. The Format Locator itself is not trainable.

The Trainable Evaluator has subfields for input locators. All of the alternatives for the input locators are tested, and the alternative with the highest confidence based on existing training documents is returned. For the best results, add five or more ideal training documents for each piece of information that you are training. If a single document contains data for multiple subfields, you do not need one for each individual subfield.

You can train the Trainable Evaluator by following these steps:

  1. Open the Project Tree window if it is not already open.
  2. Expand the Project Tree and select the class.
  3. Optionally, view the class contents if they are not already displayed.

    The hidden class contents are displayed.

  4. Open the Documents window if it is not already open.
  5. If a different view is in use, switch to the List view Documents Window - Flat View icon.
  6. Add or open a document set that contains the documents to use to train the Trainable Evaluator.

    A list of documents is displayed.

  7. Double-click a document.

    The Document Viewer is displayed showing the selected document.

  8. If the document is suitable for training, select the document in the list view of the Documents window and select Train for Extraction Train for Extraction icon from the menu.

    The Edit Document window is displayed showing the selected document.

  9. In the Edit Document window, select a field and lasso the corresponding content in the document.

    The lassoed data is entered into the field.

  10. Repeat lassoing for each field and when finished, click Add to Training Folder Add to Training Folder icon.

    The document is added to your Extraction Set and the next document in your document set is loaded automatically.

    Note If there are additional documents in the document set, these are loaded in the Edit Document window automatically, once you have processed the first document. If you do not wish to add more documents to the extraction training set for this class, close the Edit Document window.
  11. Continue to add training documents until you are ready to test your training set, and Close the Edit Document window.
  12. On the Process Ribbon tab, in the Train group, click Extraction Train - Extraction icon.

    The documents in your training set are trained and a progress bar lets you view the progress.

  13. In your test document set, right-click one or more selected documents, and click Process.

    The document is classified and extracted.

  14. Open the Extraction Results window if it is not already open.

    The extraction results are displayed. Invalid fields have a blue question mark and valid fields have a green check mark.

  15. In the Extraction Results window, view the Trainable Evaluator results based on your training documents.

    If the results are not satisfactory, add training documents to your Extraction Set by repeating steps 7 through 14.