Intent health helps guide you as you build out your intents and gives you the tools to improve more easily and accurately the performance of your intents and training utterances.

With intent health, you can refine the health status of your intents for supported individual languages. Use the view to see your intent health score based on your training data and analyze recommendations that help improve utterances that affect the health score.

The intent health view gives you key performance data, including these analysis points:

  • Intent score for each intent
  • Recommendations to improve each utterance, such as potentially confused utterances, duplicate utterances, outliers, and utterances that are too long
  • Recommendations to improve each utterance
  • The list of utterances in descending order of severity count
Note: Architect only computes health when no intent validation errors exist (for example, less than two utterances for the default language).

Health status

Architect provides the following health statuses based on NLU best practices and training data:

  • Very good
  • Good
  • Okay
  • Needs review
  • No data
  • Changes made
  • New

These statuses are meant for you to use as a guideline. Intent status indicates which intents may need improvement. Genesys recommends that you review all intents, with a focus on OK or Needs review intents.

Click the image to enlarge.

Intent health example

Utterance details from the Intents view

To train an intent, you must have a minimum of two utterances. Genesys recommends that you add at least five utterances. If you add fewer than five, the intent health returns a status of Needs review, and you can hover over the utterance to view details about it. If you add more than 20 utterances, the counter bar no longer appears, but the number of utterances remains.

Click the image to enlarge.

Suggestions to improve the utterance

For more information, see Work with intent health.