Neuro-symbolic AI and knowledge engineering
Neuro-symbolic AI (NeSy) combines the structured, interpretable reasoning of symbolic systems with the adaptive, data-driven learning of machine learning.
Symbolic systems rely onlies Knowledge Engineering (KE), i.e. the design, construction, and maintenance of formal knowledge representations such as ontologies and knowledge graphs. These provide the semantic backbone that allows artificial intelligance (AI) systems to reason explicitly, explain their decisions, and apply knowledge consistently across domains.
In NeSy, KE and neural approaches work in synergy. Machine learning can extract knowledge from unstructured data, identify patterns in large datasets, and suggest refinements to structured models, tasks that would otherwise require significant manual effort. In turn, symbolic knowledge makes neural systems more transparent, robust, and capable of reasoning beyond pattern recognition.
Our research
Our research in Neuro-Symbolic AI bridges symbolic reasoning and neural computation. Through both theoretical and translational research, we support intelligent, interpretable, and reliable systems. By uniting symbolic structures with neural models, we tackle complex, real-world challenges while keeping systems transparent and trustworthy.
In collaboration with industry partners such as Unilever, we support autonomous formulation laboratories, improving how experimental knowledge is represented, integrated, and reused. In the cultural heritage sector, we work with the British Music Experience Museum and other North-West stakeholders to build knowledge-based tools for interactive and educational museum experiences.
In the Horizon Europe project RobustifAI, we work alongside industrial partners such as Thales and Siemens to explore how neuro-symbolic AI techniques can be leveraged to develop generative AI systems that are not only more robust but also more explainable.
Our core research themes in knowledge engineering:
- Principles of ontology and knowledge graph engineering: Developing methods for creating coherent, reusable, and semantically expressive ontologies and knowledge graphs, especially within distributed environments
- Knowledge integration and reconciliation: Investigating frameworks for aligning and integrating heterogeneous ontologies and knowledge sources in distributed environments, leveraging symbolic reasoning alongside LLMs and neural representation learning to resolve entities, reconcile schemas, and construct consistent, unified knowledge models
- Automated and semi-automated knowledge acquisition: Exploring the use of generative models, statistical analysis, and neural methods to support ontology population, competency question elicitation, and semantic validation, while ensuring rigor and interpretability
- Evolution, governance, and dynamics of knowledge structures: Developing theoretical approaches to ontology and knowledge graph lifecycle management
- Evaluation and interpretability in knowledge engineering: Formulating hybrid evaluation metrics and explainable reasoning methods, and studying their application across domains such as education, industry, and cultural heritage to ensure reliable, interpretable, and practically useful knowledge representations.
Additional themes on the integration of logic and reasoning with machine learning systems:
- Formal verification of neural components: Develop hybrid verification pipelines where symbolic methods (e.g., SAT/SMT solvers) are combined with neural approximators to verify safety, fairness, or performance constraints in neural networks used in critical systems
- Large language models for computationally challenging tasks: This research explores how to leverage LLMs to approximate solutions to computationally intensive problems with improved space and time efficiency. The approximated solutions are then validated using off-the-shelf tools to ensure correctness and reliability
- Large language models for code generation: In dynamic environments, an agent’s behaviour—encoded as a program—must adapt in real time. This requires an AI agent capable of reasoning with LLMs to continuously detect environmental changes and update the generated code accordingly
- Causal reasoning with neuro-symbolic models: Incorporate causal graphs and symbolic logic into neural architectures to enable reasoning about interventions, counterfactuals, and domain shifts
- Human-in-the-loop neuro-symbolic systems: Design interactive frameworks where human experts can inject symbolic rules, constraints, or domain knowledge into neural models during inference or retraining.
Our team
- Dr. Jacopo de Berardinis - Works on multimodal knowledge graph construction and integration, metadata interoperability, and knowledge discovery
- Dr. Floriana Grasso - Her interests include argumentation-based reasoning, explainable AI, and user modelling, with a focus on how knowledge and explanations are tailored for different audiences
- Dr. Terry Payne - Works at the intersection of semantic web, multi-agent systems, and service-oriented architectures, developing frameworks that enable autonomous, distributed agents to collaborate effectively
- Dr. Valentina Tamma - Her expertise is on ontology engineering, semantic interoperability, and knowledge-based systems, with a focus on combining symbolic and neural techniques to support both theoretical and applied challenges in distributed and industrial contexts
- Prof Xiaowei Huang - Specialises on the formal verification of neural networks, uncertainty quantification of AI systems, and the logic reasoning in multiagent systems
- Dr. Reham Alharbi (honorary staff member) - Researching applications of knowledge engineering: requirement acquisition and ontology reuse in the context of cultural heritage
- Dr. Samah AlKhuzaey (honorary staff member) - Interests include semantic data integration and ontology based intelligent decision-support systems for education
- Mr. George Hannah (PhD student) - Investigates the use of large language models (LLMs) for extracting and integrating knowledge into knowledge graphs, particularly within the context of autonomous formulation laboratories.
Partnerships and collaborations
We maintain a vibrant network of academic, industrial, and cultural collaborations demonstrating both the theoretical depth and practical impact of our research.
Academic collaborations: We work with leading institutions across Europe, including the Universidad Politécnica de Madrid (Spain), Fondazione Bruno Kessler (Italy), University of Amsterdam (Netherlands), the University of St. Gallen (Switzerland), University of Oldenburg (Germany), Technical University of Vienna (Austria), and University Grenoble Alpes (France).
Industry and government collaborations: Our partnerships with industry and government showcase the real-world relevance of our work. With Unilever and the Materials Innovation Factory, we are applying knowledge engineering methods to autonomous formulation laboratories. The British Music Experience Museum collaboration has produced a knowledge-graph-driven exhibition platform, enabling semantic exploration, interactive storytelling, and innovative public engagement with the history and culture of British music.
Members are also actively involved in the Alan Turing Institute to explore Neuro-Symbolic AI and Knowledge Graphs.
Funding and support: Our reseach is supported by a range of prestigious funders, including the Engineering and Physical Sciences Research Council (EPSRC), UKRI Knowledge Transfer Partnerships (KTPs), the European Union, and the British Academy.
Opportunities
We welcome opportunities to engage with students, researchers, and industry partner:
- PhD opportunities: Please contact us if you are interested in pursuing a PhD in neuro-symbolic AI, ontology engineering, knowledge-based systems
- Consultancy and research collaborations: We welcome partnerships, consultancy, and knowledge exchange projects with academic, industry, and government partners.
Selected publications and resources
- GitHub – KE-UniLiv – Access demos, tools, and ontologies on Knowledge Engineering
- R Alharbi, V Tamma, and F Grasso. "Requirement-based methodological steps to identify ontologies for reuse." International Conference on Advanced Information Systems Engineering (EKAW). Cham: Springer Nature Switzerland, 2024
- S Alkhuzaey, F Grasso, TR Payne, V Tamma. "Text-based question difficulty prediction: A systematic review of automatic approaches." International Journal of Artificial Intelligence in Education 34.3 (2024): 862-914
- S Alkhuzaey, F Grasso, TR Payne, V Tamma. “Evaluating the Fitness of Ontologies for the Task of Question Generation.” European Artificial Intelligence Conference, 2025
- R, Alharbi, V. Tamma, F. Grasso, & TR Payne:. “The role of Generative AI in competency question retrofitting”. In European Semantic Web Conference (pp. 3-13). Cham: Springer Nature Switzerland, 2024
- T. Kampik,, A. Mansour, O. Boissier, S. Kirrane, J. Padget, T. R. Payne, M. P. Singh, V. Tamma, and A. Zimmermann. "Governance of autonomous agents on the web: challenges and opportunities." ACM Transactions on Internet Technology 22, no. 4 (2022)
- J. Chen, H. Dong, J. Hastings, E. Jiménez-Ruiz, V. López, P. Monnin, C. Pesquita, P. Škoda, and V. Tamma. "Knowledge Graphs for the Life Sciences: Recent Developments, Challenges and Opportunities." Transactions on Graph Data and Knowledge 1, no. 1 (2023): 5-1
- F. Ciroku, J. de Berardinis, J. Kim, A. Meroño-Peñuela, V. Presutti, and E. Simperl. "RevOnt: Reverse engineering of competency questions from knowledge graphs via language models." Journal of Web Semantics 82 (2024): 100822
- W. Huang, Y. Zhou, G. Jin, Y. Sun, J. Meng, F. Zhang, X. Huang. Formal verification of robustness and resilience of learning-enabled state estimation systems. Neurocomputing, Volume 585, 2024, 127643, ISSN 0925-2312
- F Wang, Y Zhang, X Yin, G Cheng, Z Fu, X Huang, W Ruan. A Black-Box Evaluation Framework for Semantic Robustness in Bird's Eye View Detection. AAAI-2025
- G Jin, R Mu, X Yi, X Huang, L Zhang. Invariant Correlation of Representation with Label. IEEE Transactions on Information Forensics and Security, 2025.
Contact us
Please discuss with relevant academic staff if you are interested in their research.