From syntax to semantics: SciLib ontology for formalization of scientific knowledge

Existing ontologies of scientific knowledge either describe publication metadata (SPAR, ORKG) or formalize specific areas of mathematics (OntoMathPRO, OMDoc), but do not provide a unified mechanism for separating semantic content, representation forms, and data carriers. This paper presents SciLib v1.0.0, a modular OWL/DL ontology implementing a three-level separation into interpretation, representation, and resource. A model of context-dependent truth is formalized through reification of the TruthAssessment class. The ontology is validated through materialization of a knowledge graph from the MathLib (Lean 4) library, including domain taxonomy construction, entity-to-class mapping, and multimodal representation generation.