OCR Confidence Score Below Threshold Word Counter
Count OCR word confidence scores that fall below a review threshold and report the affected fraction of the transcribed words.
Description
Count OCR confidence scores strictly below a configurable 0–1 review threshold and return the affected fraction.
Count OCR word confidence scores that fall below a review threshold and report the affected fraction of the transcribed words.
When to use OCR Confidence Score Below Threshold Word Counter
- Count OCR confidence scores strictly below a configurable 0–1 review threshold and return the affected fraction.
- Apply the same explicit rule consistently during ingest, quality control, reporting, or preservation planning.
- Expose records that need human review without silently rewriting the source data.
How the calculation works
A score exactly equal to the threshold is not counted; confidence calibration differs by OCR engine and model version.
fraction below = count(score < threshold)/total scores Interpreting the result
A score exactly equal to the threshold is not counted; confidence calibration differs by OCR engine and model version.
Assumptions and limitations
- The tool treats scores as independent words and does not assess layout, character errors, language-model bias, or whether confidence is calibrated to accuracy.
- Preserve the original value and record the rule version when using this result in an archival workflow.