Data formats accepted by conversational language understanding

If you're uploading your data into CLU it has to follow a specific format, use this article to learn more about accepted data formats.

Import project file format

If you're importing a project into CLU the file uploaded has to be in the following format.

{
  "projectFileVersion": "2022-10-01-preview",
  "stringIndexType": "Utf16CodeUnit",
  "metadata": {
    "projectKind": "Conversation",
    "projectName": "{PROJECT-NAME}",
    "multilingual": true,
    "description": "DESCRIPTION",
    "language": "{LANGUAGE-CODE}",
    "settings": {
            "confidenceThreshold": 0
        }
  },
  "assets": {
    "projectKind": "Conversation",
    "intents": [
      {
        "category": "intent1"
      }
    ],
    "entities": [
      {
        "category": "entity1",
        "compositionSetting": "{COMPOSITION-SETTING}",
        "list": {
          "sublists": [
            {
              "listKey": "list1",
              "synonyms": [
                {
                  "language": "{LANGUAGE-CODE}",
                  "values": [
                    "{VALUES-FOR-LIST}"
                  ]
                }
              ]
            }            
          ]
        },
        "prebuilts": [
          {
            "category": "{PREBUILT-COMPONENTS}"
          }
        ],
        "regex": {
          "expressions": [
              {
                  "regexKey": "regex1",
                  "language": "{LANGUAGE-CODE}",
                  "regexPattern": "{REGEX-PATTERN}"
              }
          ]
        },
        "requiredComponents": [
            "{REQUIRED-COMPONENTS}"
        ]
      }
    ],
    "utterances": [
      {
        "text": "utterance1",
        "intent": "intent1",
        "language": "{LANGUAGE-CODE}",
        "dataset": "{DATASET}",
        "entities": [
          {
            "category": "ENTITY1",
            "offset": 6,
            "length": 4
          }
        ]
      }
    ]
  }
}

Key Placeholder Value Example
api-version {API-VERSION} The version of the API you're calling. The value referenced here is for the latest released model version released. 2022-05-01
confidenceThreshold {CONFIDENCE-THRESHOLD} This is the threshold score below which the intent will be predicted as none intent. Values are from 0 to 1 0.7
projectName {PROJECT-NAME} The name of your project. This value is case-sensitive. EmailApp
multilingual true A boolean value that enables you to have utterances in multiple languages in your dataset and when your model is deployed you can query the model in any supported language (not necessarily included in your training documents. See Language support for more information about supported language codes. true
sublists [] Array containing sublists. Each sublist is a key and its associated values. []
compositionSetting {COMPOSITION-SETTING} Rule that defines how to manage multiple components in your entity. Options are combineComponents or separateComponents. combineComponents
synonyms [] Array containing all the synonyms synonym
language {LANGUAGE-CODE} A string specifying the language code for the utterances, synonyms, and regular expressions used in your project. If your project is a multilingual project, choose the language code of the majority of the utterances. en-us
intents [] Array containing all the intents you have in the project. These are the intents that will be classified from your utterances. []
entities [] Array containing all the entities in your project. These are the entities that will be extracted from your utterances. Every entity can have additional optional components defined with them: list, prebuilt, or regex. []
dataset {DATASET} The test set to which this utterance will go to when split before training. Learn more about data splitting here . Possible values for this field are Train and Test. Train
category The type of entity associated with the span of text specified. Entity1
offset The inclusive character position of the start of the entity. 5
length The character length of the entity. 5
listKey A normalized value for the list of synonyms to map back to in prediction. Microsoft
values {VALUES-FOR-LIST} A list of comma separated strings that will be matched exactly for extraction and map to the list key. "msft", "microsoft", "MS"
regexKey {REGEX-PATTERN} A normalized value for the regular expression to map back to in prediction. ProductPattern1
regexPattern {REGEX-PATTERN} A regular expression. ^pre
prebuilts {PREBUILT-COMPONENTS} The prebuilt components that can extract common types. You can find the list of prebuilts you can add here. Quantity.Number
requiredComponents {REQUIRED-COMPONENTS} A setting that specifies a requirement that a specific component be present to return the entity. You can learn more here. The possible values are learned, regex, list, or prebuilts "learned", "prebuilt"

Utterance file format

CLU offers the option to upload your utterance directly to the project rather than typing them in one by one. You can find this option in the data labeling page for your project.

[
    {
        "text": "{Utterance-Text}",
        "language": "{LANGUAGE-CODE}",
        "dataset": "{DATASET}",
        "intent": "{intent}",
        "entities": [
            {
                "category": "{entity}",
                "offset": 19,
                "length": 10
            }
        ]
    },
    {
        "text": "{Utterance-Text}",
        "language": "{LANGUAGE-CODE}",
        "dataset": "{DATASET}",
        "intent": "{intent}",
        "entities": [
            {
                "category": "{entity}",
                "offset": 20,
                "length": 10
            },
            {
                "category": "{entity}",
                "offset": 31,
                "length": 5
            }
        ]
    }
]

Key Placeholder Value Example
text {Utterance-Text} Your utterance text Testing
language {LANGUAGE-CODE} A string specifying the language code for the utterances used in your project. If your project is a multilingual project, choose the language code of the majority of the utterances. See Language support for more information about supported language codes. en-us
dataset {DATASET} The test set to which this utterance will go to when split before training. Learn more about data splitting here . Possible values for this field are Train and Test. Train
intent {intent} The assigned intent intent1
entity {entity} Entity to be extracted entity1
category The type of entity associated with the span of text specified. Entity1
offset The inclusive character position of the start of the text. 0
length The length of the bounding box in terms of UTF16 characters. Training only considers the data in this region. 500

Next steps

  • You can import your labeled data into your project directly. See import project for more information.
  • See the how-to article more information about labeling your data. When you're done labeling your data, you can train your model.