FAIR Assessment in CLARIAH

the interim score�

Menzo Windhouwer – Humanities Cluster Digital Infrastructure, KNAW

QiQing Ding – Humanities Cluster Digital Infrastructure, KNAW

Wilko Steinhoff - DANS, KNAW

​

metadata on FAIR tools & datasets for INEO

• tools:
• only partially available, often blended with datasets
• go get it at the source!
1. Maarten van Gompel
• requirements and guidelines
• allow us to assess the FAIRness of tools
• harvesting pipeline
• datasets:
• two major groups of partners in CLARIAH:
• CLARIN partners which are B or C centers in the european CLARIN network, which provide metadata to CLARIN's central catalogue: the Virtual Language Observatory
• Netwerk Digitaal Erfgoed (NDE) partners which provide metadata to the NDE dataset registry

​

• how to assess the FAIRness of these datasets?
• create a FAIR Implementation Profile (FIP)
• for CLARIAH as a whole: 2 years ago it was too late to do that in the timeframe of CLARIAH+
• or for CLARIN & NDE, and find their common denominators

CLARIN B Centre requirements

• CLARIN centre types
• B: meet all the (technical) requirements
• K: provides knowledge & support
• C: "just" provide metadata

​

• To become a B centre one has to:
• hand in a self assessment of the conformance to the B centre requirements
• (F) persistent identifiers
• (F) completeness
• (A) harvesting
• (A) single sign-on
• (I) component metadata
• (R) IPR
• …
• which gets reviewed by the CLARIN Assessment Committee
• next to these CLARIN specific requirements one also has to acquire a Core Trust Seal

FAIR Implementation Profile (FIP)

• together with Enno Meijers (NDE) under the guidance of Shuai Wang and Angelica Maineri (VU, ODDISSEI, SSHOC-NL) we created a first version of a FIP

CLARIAH-FIP-mini-questionaire

​

• CLARIN FIP based on B centre requirements
• top down, but once upon a time created bottom up
• NDE FIP resulted mainly in questions
• very limited number of hard requirements
• what FAIR means in NDE should come from bottom up!
• next problem: how to turn the FIP into an actual technical assessment?
• FAIR is about machine accessibility!
• there are some generic tools, e.g. F-UJI, which is linked data oriented
• suitable for NDE, which uses DCat/schema,.org (RDF) for metadata
• CLARIN uses Component Metadata (XML) as its default for metadata
• at least agree on an exchange format …
• but how to prevent we compare apples with pears across communities? 🍎 != 🍐

OSTrails: eu EOSC project started 2024

by Joaquin Lopez - WP4

FAIR assessment based on the CLARIN FIP

• teamed up with Wilko Steinhoff (DANS) to work ahead of OSTrails (Open Science Plan-Track-Assess Pathways)
• pyFAT
• assessment framework as under development by OSTrails
• express tests in a language suitable for you metadata
• XPath for XML
• SPARQL for RDF
• …
• fallback to python
• tests assess a metric (=common denominator), which is tied to a FAIR principle
• scores are combined by a community specific rubric (example)
• exchange format as under development by OSTrails

​

• the interim score
• we can assess CLARIN datasets using a partial FIP (F & A)
• not yet NDE, but we are closer to experimenting
• INEO currently still contains a mix of FAIR assessed and unknown FAIR datasets, but this will change in due course
• nothing about FAIR assessment is set in stone, its the current frontier!

​

The pipeline to INEO (for datasets)

CLARIN Partner

NDE Partner

n

n

Harvester

NDE

dataset

register

OAI

SPARQL

CLARIAH

FAIRy

Datasets

INEO

Filter & merge

RUC

1

Extract & Ingest

facets

FAIR Assessment

Lets exchange war stories 🍻

• What are datasets vs collections vs digital objects vs resources?
• How to make an OAI harvester harvest a triple store?
• Try out a JSON database?
• Sync 1000s of records with a deadline in sight
• Add RUC for your dataset(s)
• …

​

Visit us at

the INEO & FAIR demo & poster

at 15:30 in the Spot