Data validation

Checking whether data meets specified requirements before it is accepted for a use.

Validation can check structure, completeness, ranges and consistency, or assess whether the content supports a label. Passing a schema check does not prove semantic correctness. State which requirements were tested and retain unresolved items instead of treating every technically readable record as ready for training.

In practice

A timestamp can be correctly formatted but still identify the wrong moment in a demonstration.