Symbolic AI / Knowledge Graphs

Symbolic AI denotes a family of techniques that focus on the explicit representation of knowledge in formal, machine-actionable information artifacts such as taxonomies, ontologies, or knowledge graphs. As a knowledge-driven paradigm rooted in logic, symbolic AI relies on symbols to represent complex domain knowledge and intricate rules, ensuring that knowledge derivation remains completely transparent and explainable. Increasingly, these symbolic AI techniques are being used in tandem with sub-symbolic, data-driven AI—such as classical Machine Learning models, Deep Learning architectures, and Large Language Models (LLMs)—across a variety of areas, including information retrieval. Therefore, this lecture aims to provide the fundamentals necessary for understanding, constructing, and making practical use of such symbolic AI artifacts within information systems. Students will explore various knowledge organization structures categorized by their level of expressivity, progressing from basic controlled vocabularies and thesauri to structured taxonomies, heavyweight ontologies and knowledge graphs. They will also learn about formal logic-based representation languages that allow machines to parse and understand data at a meaningful level. Finally, the lecture will provide real-world examples of how symbolic AI techniques are deployed in industrial applications and research projects, offering inspiration for incorporating these approaches into contemporary information retrieval projects.