Postdoctoral researcher working on ontology embeddings, their combination with LMs, and ontology reasoning tasks. Hui will lead the sessions on geometric embeddings, LM-based embeddings, and complex ontology reasoning.
Half-day tutorial · International Semantic Web Conference 2026
OntoLM
Ontology Embedding, Reasoning and Construction with Language Models
This tutorial introduces recent work connecting OWL ontologies, ontology embeddings, and language models. It focuses on practical methods for ontology completion, inference, reasoning, alignment, and construction, with lectures, worked examples, notebook-based resources, and discussion.
Overview
OWL ontologies provide precise semantics and reliable symbolic reasoning. In real-world settings, however, knowledge is often incomplete, uncertain, noisy, and expensive to model manually. OntoLM presents machine learning and language model methods that address these limitations while keeping ontology structure central.
- Ontology embeddingsGeometric and LM-based representations that encode logical and textual information for completion and inference.
- Reasoning with LMsDeductive and abductive reasoning over OWL ontologies, including proof generation with large language models.
- Ontology constructionAlignment, subsumption prediction, and new concept placement supported by models and practical toolkits.
- Practical resourcesWorked examples, reading materials, notebooks, DeepOnto resources, and interactive discussion.
Program
The tutorial is organized into two self-contained parts: ontology embeddings, followed by ontology reasoning and construction.
| Block | Length | Content | Speaker |
|---|---|---|---|
| Part I - Ontology Embeddings | |||
| 1 | 15 mins | Introduction and roadmap | Jiaoyan Chen |
| 2 | 40 mins | Geometric embeddings: from hierarchies to axioms | Hui Yang |
| 3 | 40 mins | Language model-based embeddings | Hui Yang |
| 4 | 5 mins | Q&A and discussion | All |
| 30 mins - Conference break | |||
| Part II - Ontology Reasoning and Construction | |||
| 5 | 30 mins | Complex ontology reasoning with LLMs | Hui Yang |
| 6 | 30 mins | Ontology alignment with DeepOnto | Jiaoyan Chen |
| 7 | 30 mins | New concept placement in ontologies | Hang Dong |
| 8 | 10 mins | Summary, open challenges and discussion | Jiaoyan Chen |
Audience
The tutorial is intended for researchers and practitioners from the Semantic Web and related AI communities, especially knowledge representation and reasoning, knowledge engineering, NLP, language models, and machine learning. Expected attendance is 30-60 participants.
- PrerequisitesFoundations of ontologies, including the Web Ontology Language, plus basic knowledge of machine learning and language models.
- Learning outcomeA clear understanding of ontology embedding algorithms and how they support reasoning and construction with LMs.
Presenters
The tutorial is presented by researchers working at the intersection of ontologies, knowledge graphs, language models, and machine learning.
Senior Lecturer in Computer Science with extensive work on knowledge graphs, ontologies, and machine learning, including teaching and tutorial experience at ISWC and IJCAI.
Lecturer in Computer Science working on knowledge graphs, NLP, language models, and health informatics. Hang will lead the session on new concept placement in ontologies.
Materials
Slides, reading lists, code, datasets, notebooks, and a DeepOnto tutorial will be released openly according to the ISWC 2026 schedule.
Related Links
A compact list of papers and open-source projects corresponding to the main methods discussed in the tutorial.
- Ontology embedding survey arXiv paper
- TransBox arXiv paper
- RegD arXiv paper
- OnT papercode
- LLM4Proof papercode
- DeepOnto papercode
- BERTMap arXiv paper
- BERTSubs arXiv paper
- OWL2Vec* papercode
- Concept placement OntoLAMA paperDeepOnto code