Creating training data manually for an IVA

Ensure that Pega Intelligent Virtual Assistant (IVA) interprets chat messages from users in the correct way by manually creating training records from which the system can learn. Training records provide valuable information for the IVA that strengthens the artificial intelligence algorithms and improves the accuracy of your text analytics model. By training the model, you ensure that the system correctly detects topics and entities, the subject matter and other vital information extracted from the chat conversation. For example, you can add a training record in which customers request bank loans. If you connect that record to a bank loan topic, the system learns to associate this information with bank loan cases.
Before you begin: Enable recording of training data. For more information, see Enabling training data recording for an IVA.

You review the training data before applying the changes to the text analytics model.

  1. In the header of Dev Studio, click the name of the application, and then click Channels and interfaces.
  2. In the Current channel interfaces section, click the icon that represents your existing Unified Messaging or Web Chatbot channel.
  3. In the channel, click the Training data tab.
  4. In the toolbar, click Add records.
  5. In the Topic field in the Create new training record window, press the Down arrow key, and then select the topic to which the training record relates, for example, Bank loan.
  6. In the text field below the Topic field, enter the text for the training record.
    The text must be a good representation of sample user input, from which the IVA can learn to respond to messages.
    For example: I would like to obtain a bank loan for the amount of 10,000 dollars.
  7. Click Create record.
    Result: The system displays the training record in the list on the Training data tab. The Review training data section displays the content of the training record with the detected entities highlighted. The NLP analysis section displays the detected topic and entity types in the training data.
  8. Optional: To add more training records for the IVA, repeat steps 4 through 7.
  9. Click Save.
What to do next: Train the IVA by building a text analytics model. For more information, see Applying changes to a text analytics model for an IVA.