Unlocking the Archive: Automatic Text Recognition for Challenging Documents

This half-day workshop is jointly organized by the DICED project (Database on Intergovernmental Cooperation in Education) and the ForATR pilot project, both at the University of Zurich. DICED brings together archival records from UNESCO, the UN, and the Council of Europe into a structured, FAIR-compliant knowledge base. ForATR develops ATR-based workflows to make historical documents machine-readable and searchable. The workshop uses ATR as a connecting thread, bringing the shared experience of the two projects into dialogue with other research involving access to large collections of complex documents. The central questions are practical: What workflows and tools do projects use when Transkribus and other standard OCR/HTR implementations fall short? When, and how much, human input must remain in the loop? What is good enough for a given research question, and what potential do newer VLMs offer?