We present a framework for transforming scientific literature from unstructured text into structured, machine-readable data, enabling knowledge-driven discovery in synthetic chemistry. Our work specifically targets metal-organic polyhedra (MOPs), a class of organometallic nanocages. Building on prior rational design principles established within The World Avatar (TWA) project, we extend these efforts by focusing on chemical synthesis procedures. A dedicated synthesis ontology was developed to standardize chemical synthesis documentation, employing universal standards such as the Universal Chemical Description Language (XDL). We implemented advanced prompt engineering techniques using large language models (LLMs) to efficiently extract structured data from literature, translating ontology definitions into JSON schemas. The extracted information was integrated into the TWA ontology landscape via automated workflows. Applying this pipeline to literature, we successfully integrated data from 69 of 75 selected publications, uploading 291 synthesis procedures, 272 products, and 565 chemical species. Among these, 88 species were linked to existing concepts in the OntoSpecies ontology, and we established equivalences for 102 MOPs and 78 chemical building units (CBUs) within the OntoMOPs ontology. Selected use cases illustrate how our framework assists chemists in structured synthesis planning and documentation, demonstrating the shift from empirical synthesis toward systematic, knowledge-driven methodologies. This advancement promises enhanced efficiency, reproducibility, and innovation within synthetic chemistry.
