Keywords: Greek papyri, deep neural networks, convolutional networks, transformers, document retrieval, author classification, period classification
Project partners: Prof. Isabelle Marthot-Santaniello, Dr. Giuseppe De Gregorio
CeDA collaborator: Rodrigo C. G. Pena
Repository:

Context

Reassembling fragmented papyri, identifying authorship and dating documents are essential, albeit time-consuming, tasks for papyrologists. Computer vision and machine learning show promise in helping with these tasks, both in speed but also in implicitly providing a computable answer to the ill-posed concept of "similarity" between documents. However, machine learning approaches are often criticized for being "black boxes" that do not provide interpretable results. This project aims to develop models that both perform well in tasks related to the visual study of papyri but that are also interpretable and can be used by papyrologists to advance their research.

Project objectives

Push the state-of-the art in document retrieval, and author and period identification, while providing interpretable results to papyrologists.