Why do I get an extraction error in production?

When processing a document, certain extraction errors are detected automatically to prevent exporting incorrect data

Low Label Accuracy

When your Workflow's custom AI model's confidence in a label prediction/suggestion is low, it shows a warning in the |Production| module. Reviewing the document and confirming or correcting the AI-driven selections resolves this warning message.

Required Labels Missing

If one or more labels that have been marked as required in the Manage Labels screen are missing from the extraction, the system marks the document with the appropriate extraction error. You can resolve this error by reviewing the document and manually adding the required labels.

Data Type Mismatch

You can define the data type (e.g., Date, Integer or Float) for each label. Whenever an extracted content does not match the data type of its label, this is reported as an extraction error. Correcting this error can be done by manually changing the extracted content to match its label's format or vice-versa.

Label Validation Rule Violations

When a label validation rule is added to a label in the Manage Labels screen (e.g., "the number must at least X"), and the rule is violated, the system reports this as an extraction error. Reviewing the document and manually confirming or correcting the AI-driven suggestions is required.