Get started with conversational language understanding integration

  • Article

Important

Power Virtual Agents capabilities and features are now part of Microsoft Copilot Studio following significant investments in generative AI and enhanced integrations across Microsoft Copilot.

Some articles and screenshots may refer to Power Virtual Agents while we update documentation and training content.

This article includes the step-by-step procedures you need to get started with conversational language understanding (CLU) integration. Let’s get started.

To create a copilot using an existing CLU model:

  1. Prepare your environment.
  2. Prepare your conversational language understanding project.
  3. Map CLU intents and entities.

Prepare your environment

Before you begin, make sure that you have your CLU model project name, deployment name, Cognitive Services account key, and site URL for Azure Cognitive Service for Language.

  1. Create a new Microsoft Copilot Studio copilot.

    Note

    Ensure the environment for your copilot has been created in a region supported by Microsoft Copilot Studio and which has an equivalent CLU region.

    Not all regions supported by Microsoft Copilot Studio have an equivalent CLU region.

  2. With your copilot open, expand Settings on the side navigation pane and select Language understanding.

    Screenshot of the Microsoft Copilot Studio web app with the Settings and Language understanding options highlighted.

    Note

    If you have already enabled a CLU connection, skip to step 8.

    Note

    If you you see the following, then you are not connected to Azure Cognitive Service for Language. Follow steps 3-7 to connect.

    Language understanding option menu when not connected to Azure Cognitive Service for Language.

  3. To create a new connection, from within Power Apps, open the Connections menu by selecting ...More and then Connections.

  4. Select New connection and search for Azure Cognitive Service for Language. You are taken to an authentication settings page.

  5. Select API key for authentication type, enter account key and site URL, then select Create.

    Enter display name, authentication type, account key, and Site URL.

  6. To view the connection in the Connections list, select the Azure Cognitive Service for Language connector and select Edit in the More commands menu.

  7. Change the display name to something else to distinguish it from other CLU connectors. The name you enter to change can be any name you choose. With the connector created and configured in Power Apps, you can return to your Microsoft Copilot Studio copilot.

  8. In the Language understanding tab in Copilot Studio, choose Microsoft Copilot Studio NLU from the NLU Resource drop-down list.

    Language understanding option to select NLU resource.

  9. Choose the display name you chose earlier for the CLU connector, then select Save. A Delete example trigger phrases? window opens.

  10. Select Save snapshot. Saving a snapshot saves the copilot content, including trigger phrases and bot dialogues into a backlog file in YAML format. Select Yes, delete my trigger phrase and then select Continue.

    Select Save snapshot to delete example trigger phrases.

  11. Enter your project name and deployment information, and select Save. Choose a project and model in Azure Cognitive Service for Language.

In Microsoft Copilot Studio, on the Language understanding tab, select Save. Changing the NLU configuration removes the Analyze text system topic. It also requires you to manually add example phrases for topics currently mapped to external intents.

Now that you’re done preparing your environment, you’re ready to map intents and entities to CLU model data.

Prepare your conversational language understanding project

  1. If you don't have an Azure subscription, create a free account before you begin.

  2. Then create a language resource in Language Studio, with the conversational language understanding feature enabled.

  3. To ensure Copilot Studio functions correctly, you need to create intents for Copilot Studio system topics and for any custom topics you want to create.

  4. Author any custom entities you want to use in your Microsoft Copilot Studio project.

  5. Train and deploy the project in the same region as your Microsoft Copilot Studio resource.

Map CLU intents and entities to Microsoft Copilot Studio prebuilt data types

You create a new Microsoft Copilot Studio copilot by mapping existing topics to CLU intents. You can manually map intents and entities, or for bulk mapping, see Use the Bulk tool wizard.

Manually map intents

  1. Open your Microsoft Copilot Studio copilot.

  2. From the left navigation, select Topics and plugins and select an intent.

    Select Topics from the left navigation.

  3. In the Phrases section of the Trigger node for the intent, select the Edit link.

    Select the Edit link.

  4. In the Add phrases box, enter the name of the CLU intent that you want to map to.

    Enter CLU intent name.

    Note

    The CLU intent name must be typed exactly as stored in the CLU model, including matching case as uppercase and lowercase characters.

  5. To save the intent, click the + (plus) sign to the right of the Add phrases box.

  6. Repeat for the other intents.

Manually map entities:

  1. Open your Microsoft Copilot Studio copilot.

  2. From the left navigation, select Entities.

    Select Entities from the left navigation.

  3. Select New entity on the Entities page. Select + New entity to create a new entity.

  4. In the New entity pane, Add a name, description (optional), and Data Type for your entity. Enter the name, description, and data type for your new entity.

  5. To access CLU prebuilt entities, from the Data Type drop-down menu, select From sample data, and choose one of the CLU prebuilt JSON code snippets.

  6. (Optional) If you want to import JSON schema from a sample file, choose Get schema from sample JSON to access the external file.

Use the Bulk tool wizard

You create a new copilot by bulk mapping existing topics to CLU intents using the Map intents and entities wizard. With the wizard, your copilot can automatically utilize existing intents from your CLU model rather than trigger phrases to determine which topic to map to.

Begin by uploading your CLU model data file and automatically map CLU intents to your Microsoft Copilot Studio copilot.

  1. From the left navigation, select Language, then Add topics and entities from model data.

  2. In the File name option, select Choose file to select the data file from which you want to upload your CLU model.

  3. You then see your model data file in the Preview window. Select Next to continue.

  4. select Add intents and entities.

  5. In the Map existing topics to CLU intents screen, choose a CLU intent to map to the corresponding article. Map existing topics to CLU intents.

  6. Review your selections, then select Save. Review your selection and then select Save.

  7. (Optional) For those intents that aren't mapped to an existing topic, you can create new topics for each intent. In the wizard, you can see a list of CLU intents that are not yet mapped to a topic. In the field Create a new topic, enter a name for each of the topics that you want to create, and then select Next.

    Create new topics for CLU intents.

    Note

    Any fields that you leave blank are ignored by the wizard.

  8. Once all CLU intents and entities are mapped to Microsoft Copilot Studio topics, select Done. Add intents and entities and select Done to finish.

Once you are done with this procedure, your copilot topics now appear in the Microsoft Copilot Studio Topics pane. For more information, see Managing topics.