Extraction accuracy results

The Extraction Accuracy Details provide a multi-level accuracy view for monitoring and improving field extraction quality in a DocAI Studio solution.

At the top of this window, there are dual accuracy indicators. One for the overall solution and one for the currently selected document type, each displaying the overall accuracy percentage, document count, and the date and time of the last extraction accuracy calculation. There are also two calculation buttons to trigger fresh accuracy calculations for either scope; solution or document type.

The overall accuracy value for the solution and the selected document type represents the percentage of fields correctly extracted, regardless of the validation state, making it the primary, easy-to-interpret metric. This window helps you pinpoint which fields or document types are underperforming, so you can take corrective action.

Next, there are vertical tabs listing each document type where you can navigate between your document types. Selecting a different document type updates the field-level results for the new document type.

For the selected document type, the following information is provided.

Column Name	Description
FIELD NAME	The name of a field.
ACCURACY	The average extraction accuracy of the corresponding field. This is the sum of the CV and CI values for the field.
CV%	CORRECT AND VALID (CV) Occurs when the extracted field value matches the expected value. The goal of extraction is a 100% value, where all fields are correct and valid.
CI%	CORRECT AND INVALID (CI) Occurs when the extracted value matches the expected value but is marked as invalid. For example, a failed validation rule caused the result to be invalid. A field with this result requires manual validation during production.
IV%	INCORRECT AND VALID (IV) Occurs when the extracted value does not match the expected value, yet the field is marked as valid. This result is known as a false positive because it is mistakenly marked valid. It is not displayed to the user during production as it has a valid field status and does not require manual validation. Any results with this status need to be addressed before the solution is put into production so invalid data is not overlooked. For example, improve the extraction with new training documents that use existing fields.
II%	INCORRECT AND INVALID (II) Occurs when the extracted value does not match the expected value and the field is marked as invalid. For example, a field has no extraction result but there is a stored result, or the extracted result differs from the stored result in some other way. A field with this result in a production batch needs to be manually validated by a user.