Maciej Iwaniszewski
Reading Polish manuscripts (1689–1944) with vision-language models: an evaluation on 30 pages
Three models in seven variants on 30 manuscript pages with reference transcriptions: accuracy, cost and causes of refusals. The least expensive model achieved the highest accuracy.
The papers of Konstanty Kruszyński (1772–1818): transcription and description of 973 scans with a vision-language model
Transcription and description of 973 scans from the papers of an official who served three successive administrations (1772–1818), using a vision-language model: method, cost, reliability of readings and limitations.
Extracting publication metadata for CRIS: DOI registry, PDF parsers and language models
A comparison of a DOI registry, two PDF parsers and nine language models in extracting metadata from 960 articles in 12 languages. Registry lookup yields the highest quality; model errors concentrate in PDFs with atypical structure.
Processing museum photographs: workflow stages and the use of AI models
An analysis of the stages of museum photograph processing, from registration to publication, identifying which stages require organisational change and where AI models reduce cataloguing time.