Recent advances in foundation models have demonstrated impressive capabilities, but they still face critical limitations in reasoning, explainability, robustness, and domain adaptation. This project explores integrations of modern LMs with ontologies and knowledge graphs to address these challenges. The project will be driven by real-world domain applications in casting, materials, and manufacturing, developed in close collaboration with domain experts, ensuring both scientific depth and applied impact.
The consortium held the official kick-off meeting for the project. Both the research team and industry partners participated. The participants aligned on the project goals, collaboration model, and next steps, marking a strong start to the joint work.
We are recruiting a PostDoc Researcher in AI: Language Models, Ontologies, and Knowledge Graphs.
Bootstrapping Knowledge Acquisition from Domain Text with a Seed Ontology and Small Language Models
accepted to The 25th International Conference on Knowledge Engineering and Knowledge Management (EKAW 2026)
Iterative AI-Assisted Ontology Engineering: Lessons Learned from Developing the CASTON ontology for Metal Casting, Materials and Manufacturing
accepted to SeMatS 2026: 3rd International Workshop on Semantic Materials Science, co-located with ISWC 2026.
Ontology Refinement with Embeddings and LLMs: A Case Study on CASTON
accepted to Poster, Demo and Lightning Talk Track of 25th International Semantic Web Conference (ISWC 2026).
The work is a cooperation with Department of Computer Science, University of Manchester.
The ontology for metal casting, materials and manufacturing (CASTON)
This ontology provides a formal, machine-interpretable representation of core knowledge in casting, materials behavior, and manufacturing operations. It presents a first step toward building an AI-native semantic infrastructure for the domain.
All the serialization format for the ontology and supporting HTML documentation are openly available here.
A manuscript on CASTON is currently under the review at Scientific Data. We believe that the ontology provides a semantic backbone for potential AI applications in metal casting, materials science, and manufacturing.
From Prompt to Graph: Comparing LLM-Based Information Extraction Strategies in Domain-Specific Ontology Development
Procedia Computer Science (2026), vol 277, pp. 50–60.
Part of special issue on 7th International Conference on Industry of the Future and Smart Manufacturing (former International
Conference on Industry 4.0 and Smart Manufacturing).
The work is deveoped in cooperation with Department of Mechanical Engineering, National University of Singapore, and Department of Manufacturing Systems Engineering and Management, California State University Northridge.
Leveraging LLM and KG for Knowledge Transfer in Traditional Material Manufacturing Industry: Experience and Challenges.
SeMatS 2025: The 2nd International Workshop on Semantic Materials Science
Jönköping University
Jönköping AI Lab
Department of Computing
Department of Materials and Manufacturing
Comptech
Skillingaryd, Sweden
CONSID
Jönköping, Sweden
This project is funded by:
Grant No. 20250058
For general inquiries about the project, please contact:
He Tan
Jönköping AI lab
Department of Computing
School of Engineering
Jönköping University