Abstract
The rapid growth of research output in control engineering calls for newapproaches to structure and formalize domain knowledge. This paper brieflydescribes an LLM-supported method for semi-automated generation of formalknowledge representations that combine human readability with machineinterpretability and increased expressiveness. Based on the ImperativeRepresentation of Knowledge (PyIRK) framework, we demonstrate how languagemodels can assist in transforming natural-language descriptions andmathematical definitions (available as LaTeX source code) into a formalizedknowledge graph. As a first application we present the generation of an``interactive semantic layer'' to enhance the source documents in order tofacilitate knowledge transfer. From our perspective this contributes to thevision of easily accessible, collaborative, and verifiable knowledge bases forthe control engineering domain.