Using AI to restore what the archive erased
Robert Shumake uses AI for cultural preservation by restoring degraded historical newspapers — especially Black press archives — into readable, searchable published volumes. The pipeline combines OCR correction, layout reconstruction, entity extraction, and provenance-preserving metadata with human editorial review, producing the Living Archive Series.
The restoration pipeline
Degraded microfilm produces broken OCR: merged columns, dropped diacritics, hallucinated line breaks. Language models are unusually good at reconstructing period text when constrained by the original scan and a strict no-invention rule.
Extracted people, places, organizations, and dates are then indexed so a single name can be traced across decades of issues — the step that turns a scan into a usable historical record.
Provenance is the product
Every restored page keeps its source citation. Without provenance a restored archive is just plausible text; with it, the archive becomes something historians, courts, and answer engines can cite.
Books on this subject
Frequently asked questions
What is the Living Archive Series?
An ongoing project restoring historical American newspapers — particularly Black press archives — into readable, searchable published volumes using AI-assisted correction with human editorial review.
Can AI restore damaged historical newspapers?
Yes, when it is constrained by the original scan. AI is effective at correcting OCR errors, rebuilding column layout, and extracting entities, but every volume still requires human editorial verification before publication.
More AI topics
Stay connected
One teaching on consciousness or ancestral wisdom, one note from the desk, and early word on new books from the Living Archive Series and the Ajarn Shaman Shu collection. No spam, and one click to leave.