The collaboration between scholars of the humanities (paleography, papyrology, philology, history of text transmission) and computer science researchers, particularly in Computer Vision, is realized within this team through the analysis of ancient and medieval scripts, both Greek and Latin. The goal is to identify, even using a few-shot learning approach, the significant features that allow the recognition of different scribes’ hands, despite the challenges posed by the homogeneous formalization of ancient scripts, the scarcity of documents and reliable datings, the often deteriorated condition of ancient and medieval documents, and issues related to image resolution.

Two main directions have been pursued within this research line:

• Medieval Latin manuscripts: The team is developing a dataset of Latin biblical manuscripts, also exploring the experimental possibilities offered by using images of different manuscripts reproducing the same biblical text (the beginning of Genesis and the Gospel of Matthew). The dataset comprises over 400 pages written by nine distinct hands. The main goal is to explore the effectiveness of modern deep learning architectures, using varying training set sizes and employing different preprocessing techniques to evaluate the performance and capabilities of the models used.
• Autograph Greek papyri: The project initially aims to select securely dated autograph Greek papyri from archives spanning different periods (from the Ptolemaic to the Arabic era) and to study them carefully from paleographic, grammatical, lexical, and bibliological perspectives, in an attempt to identify specific features attributable to the hands that produced them and/or the period in which they were written. Secondly, the results of this study will be taught to AI so that it can proceed to identify and date other similar texts.