IAL Workshop with Marianne Reboul
As part of this workshop, Marianne Reboul (ENS de Lyon) and the team from the Canada Research Chair on Digital Textualities discussed the IAL (Literary Artificial Intelligence) project.
The project aims to design automated text analysis models capable of detecting literary variations in the Greek Anthology. The results from the developed algorithms are then compared to variations identified by a team of philologists. These algorithms are not developed for heuristic purposes (i.e., to discover new variations in the text), but with the idea that formalizing the concept of variation through computational algorithms can refine the philological definitions attributed to this concept.
The first part of the workshop was dedicated to an informal discussion of the following issues:
- How to treat variations that concern only part of an epigram?
- How to formally establish the difference between the concept of variation and that of a literary topos?
- How to differentiate variations from a minimal level of similarity, beyond which similarities become significant and can be associated with variations?
The second part of the workshop was dedicated to a series of experiments. The following models were specifically tested: the Ancient Greek BERT model, the multilingual alignment tool Bertalign, and the lemmatizer Stanza.