Optical Character Recognition Confidence Scores
OCR confidence scores tell you how reliably software converted scanned text from old documents, and understanding which words the software struggled with helps you catch transcription errors before they corrupt your research. Low confidence scores on names, dates, or locations warrant double-checking against the original image.
HypatiaWhen AI transcribes historical documents, it assigns a confidence score to each word or character, indicating how certain it is about the reading based on image quality, handwriting clarity, and training data. Low confidence scores flag words the AI is guessing at, which in genealogy often correspond to critical details like surnames, place names, and dates.
Understanding confidence scores helps family historians know exactly where to focus their manual review efforts, preventing errors from silently entering their research and ensuring that uncertain transcriptions do not get mistaken for confirmed facts in the family record.
Ready to work on Optical Character Recognition Confidence Scores?
Explore related journeys, or bring what you’re working through to Hypatia.