The JSON key:value pairs that describe the configuration settings for a dataset attribute.
The attributes in a unified dataset have configuration settings that indicate how the project's machine learning model acts on its data values.
The JSON object provides data about an attribute's configuration:
| KEY | Value Description | Type |
|---|---|---|
| id | The deployment-specific ID of the attribute configuration returned. Unique across the entire deployment of Tamr. Example: unify://unified-data/v1/projects/5/attributeConfigurations/57 To identify an attribute configuration in other API calls, you use the numeric value after attributeConfigurations/. | String |
| relativeId | The relative ID of the attribute configuration returned, stripped of the instance name. To identify the configuration in other API calls, you use the numeric value after attributeConfigurations/. | String |
| relativeAttributeId | The relative ID of the unified dataset attribute to which the configuration applies. | String |
| attributeRole | Optional. The specific role, if any, of the attribute in the project. Possible values are: - SUM_ATTRIBUTE ("spend" in the [user interface] (doc:working-with-the-unified-dataset#identifying-required-attributes)) - CLUSTER_NAME_ATTRIBUTE (for example, "site" or "supplier" in the user interface) Each specific role can only apply to a single attribute in a unified dataset. The default is no specific role. | String |
| similarityFunction | The similarity function to use for the unified dataset attribute. Possible values are: - COSINE - JACCARD - ABSOLUTE_DIFF - RELATIVE_DIFF Additional values apply to geospatial attributes. | String |
| enabledForML | Whether the unified dataset attribute is being included in machine learning operations. | Boolean |
| tokenizer | The tokenizer used for tokenizing text values. Possible values are: - DEFAULT - STEMMING_EN - BIGRAM - TRIGRAM - BI-WORD - REGEX Either this field or numericFieldResolution can be specified, but not both. | String |
| numericFieldResolution | Indicates how to process numeric values. Either this field or tokenizer can be specified, but not both. | Array[int] |
| attributeName | The user-defined name of the attribute. | String |
An example follows:
{
"id": "unify://unified-data/v1/projects/5/attributeConfigurations/76",
"relativeId": "projects/5/attributeConfigurations/76",
"relativeAttributeId": "datasets/111/attributes/spend",
"attributeRole": "SUM_ATTRIBUTE",
"similarityFunction": "COSINE",
"enabledForMl": false,
"tokenizer": "DEFAULT",
"numericFieldResolution": [],
"attributeName": "spend"
}