Lecture by
Martin Roček & Jan Odstrčilík
^{Faculty of Arts, Charles University & Institute for Medieval Research, ÖAW}

The use of AI in the study of medieval manuscripts and texts has been one of the most influential developments in medieval studies in the last decade. The first part of the talk will focus on handwritten/automated text recognition of historical documents, with special focus on the past and current developments at the Institute for Medieval Research of Austrian Academy of Sciences. The Particular emphasis will be given to dealing with multilingual manuscripts and related issues. Jan Odstrčilík will also present the upcoming book: Automated Text Recognition: Theory, Platforms, Best Practices by Michael Schonhardt, Tim Geelhaar, Tobias Hodel and Jan Odstrčilík.

The second part will deal with machine learning (ML) for detecting text reuse and semantic similarity in historical texts, positioning AI as a valuable assistant for researchers rather than a replacement for traditional philology. Focusing on medieval Latin dream interpretation texts (Somniale Danielis) as a case study, we address the methodological challenges of applying ML to historical languages, including semantic shifts and orthographic variations. The presentation concludes with a practical demonstration of the custom Sentence-BERT model and its application to historical texts.