A paper by Célian Ringwald, Fabien Gandon, Catherine Faron, Franck Michel and Hanna Abi Akl accepted for ESWC2025

As part of his 3IA Côte d'Azur Chair, Fabien Gandon (Inria) is honoured to have one of his papers accepted at the 22nd European Semantic Web Conference.

This paper, titled "Kastor: Fine-tuned Small Language Models for Shape-based Active Relation Extraction", is a joint project of:
▸ Fabien Gandon, 3IA Chairholder (Inria)
▸ Célian Ringwald, 3IA PhD (Inria)
▸ Catherine Faron, Head of the Wimmics research team (Inria)
▸ Franck Michel, Research engineer and researcher (Université Côte d'Azur, CNRS, Inria)
▸ Hanna Abi Akl, Professor & Researcher in Artificial Intelligence (Data ScienceTech Institute) and Adjunct Instructor (Sciences Po)

Abstract:
RDF pattern-based extraction is a compelling approach for fine-tuning small language models (SLMs) by focusing a relation extraction task on a spec- ified SHACL shape. This technique enables the development of efficient mod- els trained on limited text and RDF data. In this article, we introduce Kastor, a framework that advances this approach to meet the demands for completing and refining knowledge bases in specialized domains. Kastor reformulates the tradi- tional validation task, shifting from single SHACL shape validation to evaluating all possible combinations of properties derived from the shape. By selecting the optimal combination for each training example, the framework significantly en- hances model generalization and performance. Additionally, Kastor employs an iterative learning process to refine noisy knowledge bases, enabling the creation of robust models capable of uncovering new, relevant facts.
Keywords: Relation Extraction · Small Language Models · Structured output

ESWC25 will take place from Sunday, June 01, 2025 to Thursday, June 05, 2025 in Portorož, Slovenia.
More info