Keep the preservation source separate from replaceable derivatives
Keep an unchanged, well-documented preservation master. Create access images and OCR outputs as replaceable derivatives linked back to that master.
You should leave with distinct roles for the unchanged master, access rendition, OCR text, and layout metadata, plus identifiers that keep every derivative connected to its source.
Apply the decision with PDF to Text. Then test the tradeoff against ALTO, hOCR and Coordinate-Based OCR Formats and Sample-Based OCR Quality Assurance for Large Collections.
Use the archive OCR quality-control collection to connect master files, access copies, sampling, and measurable accuracy.
File role matrix
| Asset | Optimized for | Replaceable |
|---|---|---|
| Preservation master | Retaining source evidence | No |
| Access image | Fast viewing and delivery | Yes, from master |
| OCR text | Search, reading, data use | Yes, from source |
| Searchable PDF | Visual access plus search | Yes, from source and OCR |
Assign a role to every file in the package
The preservation master records the best stable capture and should not be replaced by a compressed viewing copy or cleaned OCR derivative. Access images support delivery. OCR text supports search and reuse. Layout files and correction records connect text back to page regions and processing history.
Use durable identifiers so every derivative names its source page. Record format, checksums where used by the repository, creation process, and relevant settings. If a derivative is regenerated, retain enough provenance to explain what changed without overwriting the original evidence.
- Separate master and derivative roles
- Connect files with durable identifiers
- Record transformations and corrections
Assign each file a documented role
Define file roles
A preservation master prioritizes retained evidence and stable format. An access copy prioritizes delivery size. OCR text prioritizes search or reuse. A searchable PDF combines a page image and text layer but should not silently replace the master.
Use durable identifiers
Assign an identifier before derivatives are made. Carry it into page filenames, OCR records, checksums, and correction logs so every output can be traced to one source.
Record transformation history
Document capture device or source, date, resolution, color mode, software, OCR engine and model, preprocessing, output schema, and later corrections. Avoid claiming exact reproducibility if versions are unknown.
Preserve fixity and separation
Use checksums and controlled storage appropriate to the collection. Keep derivatives in a structure that makes it hard to mistake a compressed access image or corrected OCR text for the original record.
Preservation package check
- Assign a persistent source identifier
- Retain an unchanged master
- Create access and OCR derivatives separately
- Record software and settings
- Link corrections to source pages
An OCR derivative is not a preservation master
OCR text, compressed access images, and searchable PDFs are replaceable outputs. They should not overwrite or become the only surviving version of the highest-fidelity source.
How preservation sources shape file roles
Institutional guidance supports separation of masters and derivatives. Local policy must still define fixity, storage, identifiers, migration, access, and the correction trail.