Digital humanities, AI, and Asian studies

Lundi 2 décembre 2024, 14h-17h | Campus Condorcet, cours des Humanités, 93300 Aubervilliers [Salle 327]

Humanités numériques, IA et études asiatiques

Demi journée d’études organisée par des membres du CCJ : Marie-Paule Hille (EHESS), Alain Arrault (EFEO) et Isabelle Sancho (CNRS).

La table ronde s’articulera autour des deux présentations suivantes :

– Cécile Armand (ENS Lyon, associée CECMC), Baptiste Blouin (Aix-Marseille Université), Christian Henriot (Aix-Marseille Université) : “From Tools to Insight: New Computational Approaches for Chinese Historical Texts in the Age of AI”

In this presentation, we share our experiences with an interdisciplinary collaboration between historians of modern China and Natural Language Processing (NLP) specialists to address the challenges of handling massive, multilingual text corpora —such as periodicals, newspapers, diaries, and directories—in studying elites in modern China. Our presentation will proceed in three parts:  1. A brief introduction to AI and an overview of how we integrate various AI techniques into our workflow. 2. A demonstration of the Corpus Creator and HistText tools we developed to build and analyze text corpora.  3. Case Studies showcasing how we have used these tools to introduce new insights into the study of elites in modern China.

– Javier Cha (Assistant Professor of Digital Humanities, The University of Hong Kong) : “Lean AI for Digital Humanities: Sustainable Innovation and Inclusive Access”

Lean AI presents a responsible, democratized vision for AI in the digital humanities, reducing dependence on Big Tech. Commercial AI solutions often lack support for historical or low-resource languages and raise privacy concerns when handling sensitive or protected data. Although open-source alternatives exist, AI-powered humanities research typically requires substantial computational power and advanced hardware. Lean AI addresses this barrier through heterogeneous clusters of repurposed or low-power systems running optimized models with minimal precision loss. By resolving technical
challenges and establishing multilingual benchmarks, we create workflows suited to diverse cultural contexts. In addition, educational initiatives and community grants promote broader adoption of a sustainable, inclusive AI ecosystem for digital humanities worldwide.