AI Digitises 350,000 Handwritten Records at Uppsala Library

Context The library’s historic “Catalog 62” listed thousands of titles in cursive script, a format inaccessible to modern search tools. Researchers and the public could only locate items by consulting the physical ledgers.

Key Facts - AI models were trained to recognise the catalog’s layout, extract author names, titles and other metadata, and output the data as XML files. - The extracted metadata were matched against the Swedish national library catalogue, Libris, and added automatically via application programming interfaces. - The process produced 350,000 searchable holdings records, now visible in Libris and the library’s own search portal. - Project manager Karin Byström says the effort equals more than ten years of manual transcription and would have been impossible without automation. - Uppsala Library is collaborating with Lund and Gothenburg university libraries, which are running parallel AI‑driven digitisation projects for their older catalogs.

What It Means Researchers can now locate rare or unique works with a few clicks, accelerating scholarly work and broadening public access. Chief Librarian Johanna Hansson notes that the initiative improves both workflow efficiency and the visibility of the collection. The partnership with Lund and Gothenburg pools expertise, creating a reusable framework for other institutions.

The project will continue into autumn 2026, creating new bibliographic records for books not yet listed in Libris. Watch for further releases of previously hidden titles as the AI pipeline expands across Sweden’s university libraries.