Terminology & Artificial Intelligence (2)

 

Generating Knowledge Graphs with LLMS, prompt engineering, and RAG

 

(including hands-on work)

 

Dates: June 3 & 4, 2025

 

Schedule: from 9:00 am to 5:00 pm (UTC+2)

 

Place: local information

 

 

 

Dr Laure Berti-Equille, Research Director, IRD French National Institute for Sustainable Development (France)
Rafail Giannadakis, University of Crete, TALOS AI for SSH (Greece)
Rachel Milio, University of Crete, TALOS AI for SSH (Greece)

 

 

Ass Prof Maria Papadopoulou, University of Crete (Greece)
Prof Christophe Roche, University of Crete (Greece), University Savoie Mont Blanc (France)

 

 

This two-day training session investigates how Knowledge Graphs (KGs) can be generated from Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG) to support terminological work in the humanities, with a particular focus on Ancient Greek texts. Building on the 2024 TOTh session on Symbolic AI, this edition combines connectionist approaches, exploring how LLMs can assist in the extraction, structuring, and enrichment of domain-specific terminology from complex historical corpora. Through practical exercises, participants will engage with Ancient Greek and Latin texts—such as treatises by Aristotle, Pseudo-Aristotle, Pseudo-Plato, and Vergil —and experiment with the use of LLMs to identify key terms, establish conceptual relations, and map them into structured knowledge graphs. Using tools like Neo4j, the LLM Graph Builder, and the Cypher query language, the session will demonstrate how these neural models can be combined with graph-based systems to enhance the transparency, traceability, and critical analysis of terminological reasoning. Special attention will be given to prompting strategies, as well as to the challenges posed by translation, ambiguity, and hallucination in generative models. By anchoring the technical exploration in the reading of ancient texts, this training aims to provide participants with both methodological insight and practical tools for bridging knowledge and modern machine reasoning.

 

 

Participants will acquire the theoretical and practical skills needed to build Retrieval-Augmented Generation (RAG) systems, leveraging Large Language Models (LLMs). Open LLMs will be used to extract concepts, entities, and relationships from texts to populate knowledge graphs.

 

 

TOTh training sessions are designed for an interdisciplinary audience of practitioners, researchers, and students: linguists, terminologists, translators, librarians, information scientists, digital humanists, cultural heritage experts, computer scientists, knowledge engineers…

 

 

 

Student: 50 €
Academic: 100 €
Other/Industrial : 150 €