Library released newspaper text tools and a bulk-data portal

The Library of Congress announced on August 11, 2026, that it had released NDNP-Open-OCR, software for improving machine-readable text from historical newspapers. The institution said it had reprocessed more than 425,000 pages and opened a datasets portal covering more than 23 million newspaper pages. ^[1]

The Library described tools for downloading entire batches, sorting records and checking archive files with SHA-256 checksums. Its software uses Python and the Tesseract recognition engine. Those are the institution’s descriptions of its tools and collection; the announcement does not establish an independently measured accuracy rate. ^[1]

The release followed an April 1 migration update that scheduled a redirect from the legacy bulk-text page to the new datasets page for April 8. That notice said bulk text would remain available, while computational researchers should move to the Library’s newer application programming interface. ^[2]

The migration also changed how older saved searches behave. The Library warned that more content and a revised search index could change results, and that it could not guarantee every link containing search terms would resolve. Its guidance was to begin a new search when needed. ^[2]

About this item: Retrospective source brief. No first-hand reporting or independent testing is claimed. Corrections policy.