Presentation for the workshop "Knowledge Organisation Systems in the Humanities - Between expert knowledge and machines" which took place at DANS-KNAW, The Hague. Decentralisation of data infrastructure is very important step and not possible without proper knowledge organisation and distributed knowledge graphs created for different disciplines. Metadata created in Dataverse data network could be harvested and stored in triple store, and queryable both by human and machine. Large Language Models could be trained to support specific data exchange standards in order to generate SPARQL queries automatically, based on the ontologies and controlled vocabularies used.