Skip to content

Latest commit

 

History

History
81 lines (67 loc) · 2.52 KB

LuisModelConfiguration.md

File metadata and controls

81 lines (67 loc) · 2.52 KB

LUIS model configuration

The train sub-command in the NLU.DevOps CLI tool accepts a --model-settings option, which allows the user to configure NLU provider-specific features such as builtin intents and phrase lists.

The --model-settings for LUIS is a pointer to a JSON file which is used as a template for the LUIS app JSON that will be used to train the model. You can use it to train either a fully defined model exported as JSON from the LUIS portal, or to combine it with a set of generic utterances supplied with the --utterances option.

Configuring LUIS features

To configure any LUIS features, such as specifying an entity type as a closed list or prebuilt entity, including a phrase list, adding built-in intents, etc., create a JSON file with the corresponding LUIS features defined and pass it to the train command via the --model-settings option:

{
  "closedLists": [
    {
      "name": "Genre",
      "subLists": [
        {
          "canonicalForm": "country",
          "list": []
        },
        {
          "canonicalForm": "hip hop",
          "list": []
        },
        {
          "canonicalForm": "jazz",
          "list": []
        }
      ]
    }
  ]
}

Training entities with roles

In order to label a LUIS role on an entity in a generic utterance, you can label the entityType using the role name, but you must also pre-define an entity type with that role in the LUIS app template provided via the train command --model-settings option.

For example, if you have the following utterance with a role-labeled entity:

{
  "text": "play two songs from my play list",
  "intent": "PlayMusic",
  "entities": [
    {
      "matchText": "two",
      "entityType": "songCount"
    }
  ]
}

Then you must also include a LUIS app template with the builtin.number entity and songCount role pre-defined:

{
  "prebuiltEntities": [
    {
      "name": "number",
      "roles": [
        "songCount"
      ]
    }
  ]
}

This will result in the labeled entity being sent to LUIS as follows:

{
  "startPos": 5,
  "endPos": 7,
  "entity": "number",
  "role": "songCount"
}

Testing entities with roles

In both LUIS v2 and v3, if a role is recognized, the entityType property will use the role name instead of the entity type.