A test case for applying a fast track approach for FDOs in the health/data science domain

The presentation was delivered by Zeyd Boukhers at the FDO Conference 2022 in Leiden. This work explores the implementation of a fast-track methodology for integrating FAIR Digital Objects (FDOs) within the health and data science sectors. It discusses a test case from the NFDI4DS project, focusing on machine learning data components such as images used for training, source code, and metadata. The presentation highlights the challenges of managing often not machine-readable metadata and proposes solutions to enhance automation in data curation and relabeling tasks. The presentation emphasizes the importance of machine-readable persistent identifiers (PIDs) and a minimal set of administrative metadata to facilitate machine-actionable decisions. It also discusses the flexibility of the design, which allows for the standardization and reuse of attributes defined in a Data Type Registry (DTR), and the potential for extending the FDO content with other metadata schemas and vocabulary specifications. It conclude by suggesting future research directions, including identifying specific attributes required for machine-actionable decisions and determining the most efficient level of data granularity for various applications