Edit a table training document

The following steps assume that you have already added one or more training documents to the Table Extraction Set for a class. When documents are first added to the Table Extraction Set they are initially excluded from training and the Exclude from Training  icon is displayed. Once a table label is added to a document it is included in training and the Include for Training  icon is displayed. This ensures that documents are excluded from training until at least one table label is added to the training document.

You can edit a table training document in the Table Extraction Set by following these steps:

Select a class and then select the Table Extraction Set.

A list of training documents that are assigned to the class are listed.
From the Documents toolbar, click Table Training .

The Table Training window is displayed with the detected table data for each document in the document set.
Use the list in the Table Training window to expand or collapse documents and view the pages with detected tables. The image viewer is updated with the selected page and table. The currently selected table is highlighted in green. All other tables are blue.

To configure a table, click on the green highlighted table in the image viewer or click Assign Table Label .

The Assign Table Label window is displayed.
Select a table model from the list and then click OK.

If the table model you need is not available, click Table Models to create a new table model.

Once the new table model has been created, you are returned to the Assign Table Label window. Select your new table model and click OK.

The label is applied to the table in the image viewer.
Optionally, once a table model is applied to a table, configure the individual columns of the table. In your newly labeled table, click on a column or click Assign Column Label .

The Assign Column Label window is displayed. The list of available columns is limited to those in the applied table model.
Select the appropriate column from the list and click OK.

If there is no appropriate column in the table model, click Table Model to edit the table model to add the column.
Repeat steps 3 to 6 for all documents in the Table Extraction Set that are assigned to this class and are set to Include for Training .
Click Table Label Settings to edit the table labels settings for your training documents.

The Table Label Settings window is displayed.
For each table label listed, select whether the tables includes a Row Header and if you want to set the Threshold value for each table label, clear the Auto optimize threshold setting and use the slider to adjust the Threshold values as needed.
Click OK to save your changes.

The Table Label Settings window is closed.
Train your table training documents by clicking on the Train and Classify Tables setting.

This training is run against the selected class and its training documents only.

The training results in the Table Training window help you see where you may need to add additional training documents. For example, if the confidence of a specific table label is consistently low, add additional training documents. Similarly, if the confidence of a specific column label is low, add additional training documents where that specific column is an ideal representation.

To run training for your entire project, use the Extraction & Table Extraction/Table Training settings from the Process tab in the ribbon. Note that when you run training against the full project, only those classes with recent changes are trained. If you have recently trained a class via the Table Training window, it is not trained unless there are changes to the training documents.
Optionally, add additional training documents.
Optionally, exclude any of the documents from training by clicking on the icon in the Use column. The Include for Training  icon indicates that the document is used and the Exclude from Training  icon indicates that the document is excluded.
When you are happy with the training results, close the Table Training window.
Optionally, train your entire project for table extraction by selecting Extraction & Table Extraction from the Process tab in the ribbon.

This trains any trainable locators as well as any untrained classes for table extraction. If you have recently trained a class via the Table Training window, that is not trained again.

Your project is trained for table detection.