Abstract
RADAR is a cross-disciplinary, FAIR-aligned repository service operated by FIZ Karlsruhe – Leibniz Institute for Information Infrastructure. It provides long-term, format-independent archiving and publication of digital research data across scientific disciplines. While datasets in RADAR are already identified by persistent identifiers and accessible via the WebUI and complete TAR downloads, the internal contents of archived datasets have long remained only partially machine-actionable. This has limited fine-grained referencing, automated reuse, and interoperability in PID-based data ecosystems. To address this, RADAR has extended its metadata and access framework to support more granular and machine-readable identification of digital objects within published datasets. Streaming-based access to individual files and folders within TAR archives (BagIt) has been available since December 2025, enabling users and machines to retrieve specific data objects without downloading entire datasets. Building on RADAR’s DOI-based identification scheme, fragment identifiers enable persistent, fine-grained referencing of files and folders within datasets. In parallel, RADAR has introduced support for the RO-Crate metadata standard. For each published dataset, an accompanying ro-crate-metadata.json is automatically generated, describing the internal dataset structure (files, directories, size, format, timestamps) alongside dataset-level metadata such as title, licence, and creators. Combined with DOI-based fragment identifiers, this establishes a machine-actionable layer in which dataset components become precisely identifiable, citable, and directly accessible via PID resolution. Building on this machine-actionable access layer, RADAR integrates the Terminology Services for NFDI (TS4NFDI) to strengthen the semantic interoperability of PID-identified objects. In an initial implementation, the metadata field “keywords” is linked to standardised terminologies and semantic artefacts provided via TS4NFDI. This supports consistent annotation, improves interpretability across disciplines, and facilitates integration into knowledge graphs and other semantic infrastructures . The approach establishes a foundation for extending terminology-based annotation to other RADAR metadata fields, e. g. subject area or rights information. Complementing these developments, RADAR explores AI-supported services for metadata enrichment and FAIRness assessment, including automated keyword extraction from descriptive metadata and text-based dataset contents. These approaches are currently in an exploratory phase but demonstrate strong potential to enhance metadata quality and support users in improving the FAIRness of their datasets. Together, these developments demonstrate the value of persistent identifiers for all stakeholders: enabling granular citation for researchers, efficient data access for machines and applications, and improved interoperability for infrastructures. RADAR thereby strengthens its role within NFDI, EOSC, and the evolving FAIR Digital Object ecosystem, contributing to keeping knowledge connected through PID-based, machine-actionable research data. The poster is being presented on occassion of the PIDfest 2026, Leiden/NL (27– 29 October 2026)