SemSys at Semantics 2025

Last week, our team attended SEMANTiCS 2025, where researchers, industry experts, and business leaders explore trends and applications in Semantic Web, Machine Learning, Data Science, Linked Data, and Natural Language Processing. The event provided a fantastic opportunity to learn about the latest advances in semantic technologies and connect with the broader research and industrial community.

We were excited to have Alex, Fajar, Gregor, Katrin, Majlinda, Marta, and Stefani representing our team, with Alex, Katrin, Majlinda, and Stefani also volunteering their time to help make the conference run smoothly.

Workshop Presentations and Organization:

• Majlinda presented her work at the KG-NeSy Workshop, sharing insights on combining knowledge graphs with neuro-symbolic AI.

• Katrin presented at the SentIS Workshop, on Data-Driven Augmentation of Expert Causal Knowledge in Cyber-Physical Systems

• Fajar, Katrin, and Marta played key roles in organizing the KG-NeSy and SentIS workshops, helping make these sessions successful and engaging for all participants.
• Marta and Fajar served as chairs for the Querying Knowledge Graphs session and KG-NeSy workshop, respectively.

The conference also featured a series of inspiring keynote and invited talks from leading experts, offering fresh perspectives on the future of semantic AI and its real-world applications. Prof. Dr. Hannah Bast opened with “The QLever SPARQL Engine,” demonstrating how performance can vary widely across SPARQL engines and highlighting the design choices that make QLever excel, especially for large knowledge graphs with billions of triples. 

Prof. Dr. Heng Ji followed with “Never-Ending Knowledge Acquisition with and for Large Language Model Agents,” where she introduced WINELL, a multi-agent framework for continuously updating Wikipedia, and discussed how LLMs can remain accurate, up-to-date, and fair.

In his talk “Psychoanalysis (and Therapy) of LLMs”, Prof. Dr. Georg Gottlob examined both the strengths and shortcomings of LLMs, identified common causes of errors, and presented the Chat2Data project as a step toward more reliable knowledge curation.

Dr. Felix Sasaki then spoke on “Semantic Modeling for Scaling AI Assistants and AI Agents in the Enterprise”, emphasizing how semantic modeling enhances explainability, efficiency, and grounding in enterprise AI systems.

Finally, Prof. Dr.-Ing. Michael Färber delivered “LLMs and Knowledge Graphs for Science: From Papers to Insights”, showing how the combination of knowledge graphs and LLMs can transform vast scientific literature into structured, evidence-based insights.

SEMANTiCS 2025 was a wonderful chance to exchange ideas, showcase our work, and strengthen our connections with the global community. We look forward to applying these insights in our ongoing projects!