Big Books, Big Data: a Metadata and HTR Approach to Three-Column Medieval Manuscripts

Can Transkribus identify a medieval scribe the way forensics identifies a fingerprint? This project tests exactly that idea, training hand-specific HTR models on three-column manuscripts — a format that signals professional book production — and then using CER proximity scores to attribute new material to known scribes. The "scribal fingerprint" concept offers a new computational approach to one of palaeography's oldest questions: who wrote this, and what else did they write?