AI Turns Uppsala’s Handwritten Catalog into Searchable Records

Context The library’s “Catalog 62” has long been a barrier for researchers because its entries exist only in handwritten form. Traditional transcription would require staff to read and type each line, a task estimated at more than ten years of full‑time effort. By training an artificial‑intelligence model to recognize the catalog’s layout and script, the library automated the extraction of author and title metadata.

Key Facts - The AI system produced an XML file that separates metadata by region, then matched each entry with existing records in Libris via application‑programming interfaces (APIs). - The resulting 350,000 holdings are now searchable both in Libris and the library’s own catalogue, streamlining the path from query to order. - Project manager Karin Byström said the effort “equals over ten years of manual work and could not have been done by hand.” - Collaboration with Lund and Gothenburg university libraries helped refine the method and share expertise. - Chief Librarian Johanna Hansson highlighted the project’s impact on research visibility and accessibility. - Work will resume in autumn 2026 to create new bibliographic records for books still missing from Libris, extending digital access to rare and unique titles.

What It Means The AI‑driven digitization removes a major bottleneck for scholars, allowing instant discovery of works that were previously hidden in handwritten ledgers. It also demonstrates a scalable model for other institutions facing legacy catalogues. As the autumn 2026 phase adds fresh records, the library’s digital footprint will expand, offering a richer pool of primary sources for future research.

*Watch for the rollout of the new bibliographic records later this year and for similar AI projects at other European libraries.*