FAIR Training Materials for Disciplinary Research Infrastructures: Metadata, Vocabularies, and Selective Harvesting in the H2IOSC Ecosystem

In a lifelong learning society and especially now in the era of Artificial Intelligence, high-quality, flexible learning resources are essential for acquiring the new skills required by the ever-evolving job market. Research infrastructures play a central role in providing users with specialised training and advanced digital tools, fostering innovation in their fields. CLARIN ERIC, the Common Language Resources and Technology Infrastructure, has consistently promoted training initiatives in linguistic and language technologies, especially through the CLARIN Learning Hub. The Italian national consortium CLARIN-IT has developed this approach within the NRRP project H2IOSC, aiming to create a cluster of research infrastructures in the field of Social Sciences and Humanities. In this context, CLARIN-IT led the work package dedicated to training and developed two platforms: an e-learning portal and a digital library for long-term deposit and curation of training materials as FAIR digital objects. The Skills4EOSC FAIR-by-Design methodology and the SSHOC vocabularies adopted in the H2IOSC project ensured metadata compatibility and interoperability of learning resources across domains. Moreover, the adopted standards will enable harvesting resources in the H2IOSC Marketplace and selective harvesting for domain-specific platforms such as the CLARIN Learning Resource Catalogue. This paper proposes an architecture for metadata mapping and harvesting of training resources to maximise their reusability across infrastructures.