Value Classifications

Value classifications is the ability to map the data values themselves through classification, independent of the property names. For example, you may want to create the below value mapping:

Source Data

{
    "size": "Small"
},
{
    "size": "Medium"
},
{
    "size": "Large"
}

Mapped Data

// note that the property name can be different or the same.
// Lume will do the property mappings + value mapping together, if applicable
{
    "product_size": "S"
},
{
    "product_size": "M"
},
{
    "product_size": "L"
}

How to classify values?

To map values via classification, leverage the enum property in your pipeline’s target schema. Enums allow you to use AI to classify data into your internal enum definitions.

For example, if your internal models work with product sizes as “S”, “M”, “L”, and you desire for all incoming data to map to one of these values, define the target property like. below. Any source data provided then will be classified into this enum, if the AI finds source data for product size (e.g. successfully maps the property).

    "product_size": {
        "description": "The size of the shirt product",
        "type": "string",
        "enum": ["S", "M", "L"]
    }

Classification rate limiting

Classification tasks require heavier AI compute, and thus hits global rate limits faster. For each job, keep the number of classifications (calculated by number of records x number of enum target properties) below 500. As a reference, the above example would trigger 3 classifications (3 records x one enum target property)

Editing Classifications

Users can now edit classifications and generated lookup tables via the API and the Lume Webapp. See Edit Mapper Guide to learn more.