RO-Crate¶
By representing the entirety of a digital object in an extensible linked-data format, RO-Crate allows digital objects to act as dynamic components in scientific workflows. An extensive toolchain supporting automation and validation enables RO-Crates to be integrated into existing workflows.
RO-Crate is widely used in a variety of domains with applications including:
Gas chromatography–mass spectrometry of samples returned from the asteroid Bennu by the OSIRIS-REx mission.
Tracking workflow provenance in the Barcelona Supercomputing Center’s distributed computing framework.
Automating lab processes at the Hopkins Extreme Materials Institute.
Describing complex climate datasets in the Coupled Model Intercomparison Project.
Enabling AI-ready processes and datasets in the UK’s Physical Sciences Data Infrastructure.
RO-Crate enables actionable data by supporting domain-specific ontologies, records detailed retrospective provenance of experimental processes, and supports data-stewardship by enabling interoperable and verifiable metadata.
Cryptographic Methods for FDO Verification¶
Trust, both in the provenance of datasets and also in the accuracy of descriptions of the processes that operate on those datasets, plays an important role in establishing confidence in FAIR Digital Objects.
But discussions around practices for establishing trust in FDOs often focus on policies intended to encourage qualitative measures of trust: “They’re a prestigious institution”, “We’ve known the PI for years”, “They publish in the best journals”, etc.
This approach neglects the many technical mechanisms which can be used to make provable and definitive assertions about FDOs: digital signatures, Message Authentication Codes, Public-key cryptography, Merkle logs, zero-knowledge proofs, Byzantine fault tolerance, and other cryptographic methods are available to provide unfalsifiable provenance records for FAIR Digital Objects which are provably correct rather than relying on vague notions of “trust”.
Scientific Workflow Automation¶
We can have nice metadata when someone types it into a box on a webpage, but that approach doesn’t scale.
Automatic annotation of instruments, datasets, experiments, processes, workflows and their outputs allows for scalable, interoperable, and machine-actionable FDOs.
Numerical Measures of Data Quality¶
A component of the CDIF4EOSC project is to extend the Cross-Domain Interoperability Framework to include a Data Quality profile. I participate in this effort, with a focus on Climate/Earth Science/Geospatial data and in particular how quantitative measures of data quality can be established which are sufficiently general to be applicable across scientific domains, yet descriptive enough to make meaningful claims about specific types of dataset.