They may be use cases where you would like to combine structured and unstructured data to enhance model performance. Categorical data is the first data type Graft is supporting and in combination with unstructured data can be used to generate similarity searches or predictions.
Categorical data is a data set which can be broken down into groups, examples include:
- Days of the week: mon, tues, wed etc
- Months of the Year: jan, feb, mar, etc
- Color: red, blue, green etc
- Ranges - such as Age (<18, 19-30, 31-49), Price (Low, med, high)
Categorical fields can be embedded in a similar manner to other text fields and used to perform similarity searches from the generated embeddings, and can be used independently or concatenated with other embedding results within the same Entity.
COMPLETE CATEGORY DATA
When ingesting categorical data each row MUST contain a value. Empty/Null values are NOT Supported at this time