Skip to main content
The documents we are going to process can be divided into two classes based on the layout. We want to create a separate set of extraction rules for each class. We also want our skill to be trainable using standard Vantage technologies. Let’s configure the document processing flow of the skill according to these requirements.
  1. Go to the Activities tab.
  2. First, we need to determine the document class. To do this, we’re going to use a classification activity. Click the Classify By Text and Image activity on the Activities pane. It will be added to the workflow. When you add this activity, a new field will be created in the skill structure. This field will be used to record the classification results and will be displayed in the skill field structure, however, it will be marked as a hidden field and will not be editable. You can change the name of this field in the Activity Properties pane on the Activities tab. Rename the field to “Layout”.
  3. Depending on the classification results, documents should be routed to different activities inside the processing flow. To fork the workflow, insert the IF activity: a. Click on the Classify By Text and Image activity in the workflow. b. In the pop-up window, select the IF item. It will be added to the workflow after the Classify By Text and Image activity.
Note: Instead of using the IF activity, you can create a workflow item containing two Extraction Rules activities and select the “Layout” field as a source field so that its values serve as conditions for choosing which activity will be applied to each document. For more information about this option, see Several sets of Extraction Rules within a single activity.
  1. Now we’ll add the activities for extracting data from documents of different classes. a. Select the Extraction Rules activity as the next item. Rename it to “Sick Note DE”. This activity will extract data from the German documents. b. Now we need an activity that will extract data from Belgian and Dutch documents. There are variants inside this class of documents, but they can be handled by a single Extraction Rules activity. Click the IF activity and add another Extraction Rules activity to the workflow. Rename the activity to “Sick Note BE-NL”. The IF activity should now be connected with both Extraction Rules activities and with Finish.
  2. We want Vantage to further train the skill using the processed documents. For this we need to add the Fast Learning activity at the end of the workflow. We will also train this activity when configuring the skill, but the main idea is to enable Online Learning for the skill during production. a. Click on “Sick Note BE-NL” in the processing flow. b. In the pop-up window, select the Fast Learning activity. It will be added after the “Sick Note BE-NL” activity, but the “Sick Note DE” activity will still be connected to Finish. c. Hover the mouse over the arrow that connects the “Sick Note DE” activity and Finish. The arrow should turn orange. d. Drag and drop this arrow on the Fast Learning activity. e. Hover the mouse over the arrow that connects the IF activity and Finish. Drag and drop this arrow on the Fast Learning activity as well. Now all documents will be forwarded to the Fast Learning activity.
Your final setup should look like this: Workflow diagram showing the processing pipeline The exclamation marks indicate that the activities haven’t yet been configured. We’ll configure them in the next steps.