ORION-Driven Research on AI-Based Semantic Annotation of Energy Data Presented at IJCKG 2025

Held from 15 to 17 October 2025 at the Foundation for Research and Technology – Hellas (FORTH) in Heraklion, Greece, the 14th International Joint Conference on Knowledge Graphs (IJCKG 2025) brought together the international research community working on Knowledge Graphs, the Semantic Web, Linked Data, and Artificial Intelligence. 

The ORION project was represented at the conference by Zhiyu Pan and Fengshuo Hao from RWTH Aachen University, Germany, who presented the paper “Semantic Annotation of Energy Data Using Ensemble Decision-Making with Large Language Models”. The paper is authored by Zhiyu Pan, Yongli Mou, Fengshuo Hao, Yuting Gao, Stefan Decker, and Antonello Monti.

The conference was attended mainly by academic researchers and PhD students, together with applied researchers and practitioners working on semantic technologies, knowledge graphs, and AI-based data integration. According to Fengshuo Hao, the experience was very positive and intellectually stimulating, highlighting the high quality of the paper presentations, the in-depth technical discussions, and the opportunity to exchange ideas with researchers working on semantic interoperability, ontology engineering, and AI-assisted knowledge graph construction. The conference also provided valuable feedback on the presented work and helped identify current research trends in the field. 

The presented paper addresses data interoperability challenges in the energy sector, where similar concepts are often described differently across datasets. The proposed approach combines Large Language Models with ensemble decision-making and ontology-based reasoning to automate and improve semantic annotation processes. The method improves both accuracy and robustness and was validated using real industrial energy datasets.

This work contributes directly to the objectives of the ORION project, particularly in supporting data and model interoperability. The ORION project was acknowledged in the paper and referenced during the presentation as the broader research framework supporting this work.

Further details and access to the full paper will be shared soon. Stay tuned!