Generative AI and Knowledge-Based Models

Second cycle of study, master’s degree

The participants:
• can explain knowledge representations (knowledge graphs, ontologies, taxonomies, rule systems) and analyze their areas of application
• can design hybrid AI architectures and connect GenAI with symbolic and structured knowledge models
• can plan, develop, and evaluate RAG and KGQA systems
• can design ontologies and knowledge graphs as the structural backbone for GenAI-based applications
• can develop prototype solutions

not applicable

- Introduction to knowledge-based AI: forms of representation, ontologies, knowledge graphs
- Challenges of statistical models (hallucination, lack of consistency, context limitations)
- Retrieval-Augmented Generation (RAG): architectures, indexing, embeddings, pipelines
- Knowledge graphs: construction, querying (SPARQL), reasoning, semantic technologies
- Hybrid AI approaches: integrating GenAI, classical AI methods, and symbolic reasoning
- Ontology-guided prompting and constraint-based prompting
- Evaluation methods: knowledge grounding, fact-checking, consistency metrics

- Russell, S., Norving, P. (2021): Artificial Intelligence: A Modern Approach
- Allemang, D., Hendler, J. (2020): Semantic Web for the Working Ontologist

Portfolio exam

German

6

Lecture, group work, presentation, and discussion of assignments

3