The Canadian Research Chair in Digital Textualities @ Canadian Society for Digital Humanities
The Canada Research Chair in Digital Writing team will be present at the annual conference of the Canadian Society for the Digital Humanities CSDH/SCHN in Toronto from May 30 to June 1. Several presentations are scheduled during the conference:
Pour des implémentations critiques de l’IA en SHS. Peut-on résister au monopole Big Tech ?
Speakers: Marcello Vitali-Rosati, Louis-Olivier Brassard, Victor Chaix, Frederic Clavert, Roch Delannay, Giulia Ferretti, Clara Grometto, Servanne Monjour, Nicolas Sauret, Alexia Schneider
L'émergence des LLMs (Large Language Models) et leur utilisation croissante, notamment en contexte universitaire, ouvrent des portes autant qu'elles soulèvent des questions en termes d'éthique et de méthodologie scientifique. Leur bon usage dans le cadre de la recherche dépend d'une implémentation adaptée prenant en compte les besoins d'explicabilité, de transparence, d'ouverture ; une implémentation propice à une compréhension et une analyse critique de leur application.
Ce panel propose de questionner l'usage des LLMs dans le cadre de la recherche en sciences humaines, en particulier ses implémentations dans des environnements d'écriture et de recherche documentaire. Nous allons présenter une série de réflexions théoriques et de projets pilotes developpés dans le cadre du partenariat CRSH Revue 3.0.
Intelligence Artificielle Littéraire - Ancient Greek and New Models
Speakers: Yann Audin, William Bouchard, Mathilde, Verstraete, Marcello Vitali-Rosati
This presentation covers advances on Project IAL - Intelligence artificielle littéraire/Literary Artificial Intelligence which is moving into its second phase. The main goal of IAL is the formalisation of literary definition, which we tried to achieve using classical methods in natural language processing and data mining. However, the results of the first phase made it obvious that the studied concept – epigrammatic variation in the Greek Anthology – involves modelling aspects of the text that go beyond its formal features. We thus introduce the use of Large Language Models trained on Ancient Greek texts to detect synonyms and other literary devices such as narrative structures. We further evaluate our current model on lemmatized Ancient Greek text with the Stanza library developped by StanfordNLP.
Computer-Assisted Hermeneutics of Philip K. Dick
Speaker: Yann Audin
In this presentation, Yann Audin unveils a new tool for literary studies. It transforms a literary corpus into a hypertextual database inspired by wikis. This first version uses the SpaCy Python library to recognize named entities, and free applications such as Obsidian and Zettlr to browse the texts of a corpus using cards that group together the use of proper nouns, places, organizations, products, etc. in the corpus under study.
Algorithmic Re-Conciliations Between Form and Matter : An Epistemological Reflection
Intervenante: Giulia Ferretti