Nowadays, Artificial Intelligence is revealing itself as an essential tool for many tasks pertaining to any kind of field. When it comes to historical studies, AI might be trained for the purpose of automatic recognition of texts. Pursuing such a goal, Isabelle Marthot-Santaniello applied Deep Learning-based methodologies to papyri. In the D-scribes project she worked towards identifying all the scribes responsible for the notarial documents of the archive of Dioscoros of Aphrodito. To do so, first and foremost she established a ground truth, consisting of a preliminary dataset of images representing the known handwriting of each scribe. These samples were narrowed down to καὶ-s along with some single letters (ε, κ, μ, ω), which display marked palaeographic features and show few, if any, variations in their ductus. With these specimina a confusion matrix was produced, a table that is used to define the performance of a classification algorithm. Currently, this modus operandi works well with inter-writer discrimination, but struggles with intra-writer variations. The same principle was applied to dating papyri.
Editors’ Choice: Report on the Erasmus+ Blended Intensive Program (BIP) “Intensive ENCODE: Digital Competences in Ancient Writing Cultures”
