AI and heritage: "artificial pasts"
- Raised
- $0
- Goal
- $100,000
Or, How Philology and Artificial Intelligence Can Serve Burned Manuscripts
Recent advances in philological tools and technologies, augmented data and artificial intelligence can help reconstruct manuscripts that have been partially or totally destroyed by fire.
These are skills cultivated at the École des chartes, notably through our courses in the Master’s program in Digital Humanities, and they are one of the research themes pursued by our faculty. We are able to generate fakes in order to train automatic transcription models and to visualize probable reconstructions. We now wish to work on actual burned manuscripts and attempt to restore their original appearance.
This work would make it possible to visualize one possible version of the original manuscript and to engage our students in the study of forgery and its detection. It would also allow us to develop expertise for discussing fakes more broadly: detection, source reliability, and critical examination of results.
Possible object of study: Old French manuscripts held by the National and University Library of Turin, which suffered a fire in 1904 and bombings in 1942 that destroyed half of its manuscripts (more than 30,000 volumes) and damaged the other half.
Funding request: $100,000, covering one year of a research engineer specializing in machine learning, publication costs, and travel to conferences.