Thank you for your feedback.
Form temporarily unavailable. Please try again or contact docfeedback@servicenow.com to submit your comments.

Train and test your NLU model

Log in to subscribe to topics and get notified when content changes.

Train and test your NLU model

Train and test your model iteratively so that its intents and entities are validated and compiled, and your model is assigned a version number.

Before you begin

  • Make sure that the NLU Model Builder - Core plugin, NLU Model Builder plugin, and Predictive Intelligence plugin are all installed and activated.
  • Create an NLU model.
  • Create one or more NLU intents and their associated entities.
  • Role required: Admin or Delegated Developer role (with permission of All File Types)

About this task

In this example scenario, you've already created and trained numerous intents, utterances, entities, and their associated annotations. In this example procedure, you're testing the NLU model by providing the system with utterances so it can deliver prediction results and confidence scores.

Procedure

  1. Navigate to System Applications > Studio.
  2. In the Application Explorer, navigate to Natural Language Understanding > NLU Models > NLU for Access Requests.
    NLU for Access Requests is the name you saved in the system when you created your NLU model. See Create an NLU model.
  3. Make sure the objects and data you've created are complete in the NLU Model, NLU Intent, and NLU Entity screens. If key data is missing or incomplete, remedy the situation.
  4. In the Intents section of the NLU Model screen, click Train.
    The system validates and compiles any changes you made to the NLU model. If the training doesn't succeed, an error message appears with guidance for resolution.
  5. Click Test.
  6. In the Test Model panel, enter an utterance or partial utterance from your utterance examples.
  7. Click Go.
    The system predicts the top intents and entities and shows you their matching confidence scores. In this scenario, you enter reset my password to a@b.com. That utterance result has a 94% match (confidence score) to the utterances you provided in your Reset Password intent.
    Shows an image of the Test Model pane where the system predicts the top intents and entities and shows you their matching confidence scores.
  8. Click Train.
  9. Click Save.

What to do next

  • If you want to update the current Confidence Threshold(%) value, click the Properties button in your NLU for Access Requests NLU Model screen, set a new value, and click Save.
  • Publish your NLU model.
Feedback