Separate layers
Source metadata remains intact. OCR, captions, entities, labels, and confidence signals are stored as derived information with clear provenance.
Generated fields can make a collection easier to use, provided they remain distinguishable from the institutional record and can be checked.
Source metadata remains intact. OCR, captions, entities, labels, and confidence signals are stored as derived information with clear provenance.
The workflow can support inspection, correction, approval, or rejection before derived fields are reused in search, research, or publishing.
We identify where a method is reliable enough to help and where ambiguity, rights, or local knowledge require human judgment.
Each project starts with the data, the context, and the intended use.
[ Discuss your project ]