Updating the NLP model for an Email channel

You can update the natural language processing (NLP) model for a configured Email channel by reviewing record items that were saved by Pega Intelligent Virtual Assistant for Email. Each record item in the list represents a received email message from a customer that is saved by the Email channel. Each record item has an assigned category that is used for text analysis. Only record items that generate no match or multiple matches text analyzer warnings are saved and appear in the Training data tab.

The NLP model for the configured Email channel is saved as a text analyzer rule. The Pega Intelligent Virtual Assistant for Email uses this rule to analyze the text of the received email for sentiment analysis, text (category) classification, intent analysis, and entity extraction.

By updating the NLP model for the configured Email channel, you support the machine learning capability for the Email channel instance. You send feedback in the Training data tab on the outcome of text analysis. When you edit a record item and match it with the expected outcomes, the subsequent text analysis results have a better confidence score, and the sentiment analysis, text (category) classification, intent analysis, and entity extraction are more accurate.

To get the training data added to the NLP model for the Email channel, a designer or a data scientist must open the Analytics Center portal ( pyDecisionAnalytics ), select the taxonomy for the Email channel, and build the model with the updated training data. For more information, see Analytics Center portal.

Note: You must purchase a separate license before using Pega Intelligent Virtual Assistant for Email in your application.

To update the NLP model:

  1. Open an Email channel interface.
  2. Click the Training data tab. A list of record items that were not yet reviewed, is displayed.
  3. Edit a record item.
    1. Optional: In the Category list, select a category name.
    2. Optional: Edit the record item content.
      1. Click the Edit icon.
      2. In the Update text field, edit the text for the received email.
      3. Click OK.
  4. Optional: Repeat step 3 to edit additional record items.
  5. Select the record items for which you want to update the NLP model.
  6. Click Update model with selected items.