Data FAIRness Practice

Dr Martin Uhrin
Dr Noel Vizcaino MIET IEEE

Synopsis

• Accessible PIDs: graph structure preserved
• Parsable metadata payload (fetched)
• List of expected good practices
• Use international standards

High level FAIR summary ...

Findable

• All IDs are global (eg UUID, hash)
• All IDs are permanent IDs (PIDs)
• Belong to a formal PID scheme
• Self ID: Dataset and Distribution (files)
• Can metadata be retrieved?
• Serialisation parsable?
• It includes core elements for findability e.g. creator
• Optional multiple online sources

Accesible

• Access level
• Metadata and data accessible by standardised protocols
• Separate access for data and metadata

Interoperable

• Using formal knowledge representation language
• Using semantic resources (eg namespaces)
• Metadata and data formally associated

Reusable

• Metadata for files? (distribution)
• Licensing info
• Using semantic resources (eg namespaces)
• Attribution/Provenance processable (PROV-O or PAV)
• Per community, metadata standards (list)
• (List) Open standards for files (eg CSV) or long term (ISO/TR 22299) or scientific formats
  

FAIR analysis and scoring

• Take results/score with a grain of salt!
• QA automation (aiding) should be the goal.
• Accounting only deployed data/scenarios
• Not progressive (FAIR/ER). And cascading failure on any issue.
• Note tools checking PIDs and other items by string methods and/or regexing.
• More semantic resources is not necessarily better. Harcoded Sets.
• Checking the unkown (absurd requirement), scope problems (software engineering).
• Automation is challenging
• Different tools have different aims.
• DCAT or schema.org serialised as any RDF FAIR by design (now much?): issue detectors

Some FAIR assessment tools (somewhat related)

• F-UJI
• FOOPS!
• O'FAIRE

Exercise options

• Make a simple data example FAIRer
• Make DCAT dataset in JSON-LD
• Must validate in playground

The End