IMC 2026: Sessions
Session 1335: Digital Medieval Studies, IV: Machine Learning, Reliability, and Evaluation
Wednesday 8 July 2026, 16:30-18:00
| Sponsor: | Digital Medievalist / Digital Humanities, University of Southampton |
|---|---|
| Organisers: | Delphine Demelas, Digital Humanities Research Institute, University of Southampton N. Kıvılcım Yavuz, Institute for Medieval Studies / School of History, University of Leeds |
| Moderator/Chair: | Delphine Demelas, Digital Humanities Research Institute, University of Southampton |
| Paper 1335-a | Auditing Datasets for Automated Handwritten Text Recognition with CERatosaurus (Language: English) Index terms: Computing in Medieval Studies, Manuscripts and Palaeography |
| Paper 1335-b | Transkribus Model Training for 11th- to 12th-Century Cyrillic Codices: Challenges and Possibilities (Language: English) Index terms: Computing in Medieval Studies, Language and Literature - Other, Manuscripts and Palaeography |
| Paper 1335-c | Large Language Models in the Service of Medieval Text Integration: Wenzelsbibel Meets MHDBDB (Mittelhochdeutsche Begriffsdatenbank) (Language: English) Index terms: Computing in Medieval Studies, Language and Literature - German |
| Abstract | The Digital Medieval Studies sessions bring together scholars interested in applying computational tools to medieval studies. The final session addresses the essential yet often overlooked foundation of reliable digital scholarship: ensuring that automated tools function accurately and earn the trust of the scholarly community. Odstrčilík and Roček introduce CERatosaurus, a sophisticated tool for auditing handwritten text recognition datasets to identify and correct errors that could compromise the reliability of automated transcription models. Kalashnikova develops systematic preprocessing methods for training accurate handwritten text recognition models in 11th-12th century East Slavic Cyrillic manuscripts, showcasing unique challenges of these medieval book layouts and abbreviation systems. Hintersteiner evaluates the methodological challenges, potential biases, and sustainability issues involved in integrating large language models into medieval German textual scholarship, comparing commercial versus open source approaches. Paper 1 is co-authored by Martin Roček, Institut für Mittelalterforschung, Österreichische Akademie der Wissenschaften |