Abstract
Semantic artefacts (i.e. ontologies, controlled vocabularies, thesauri, metadata schemata,...) and rich metadata are the cornerstone of the FAIR principles (Wilkinson et al., 2016). The principle I2 actually states that semantic artefacts should be themselves FAIR. To make such artefacts FAIR, the FAIRsFAIR project proposed 17 generic recommendations and 12 best practices integrating existing communities practices (Le Franc Y. et al., 2020, Hugo W. et al. 2021, Le Franc et al., 2022). Other complementary best practices (Garijo et al., 2020; Cox et al., 2021) have been proposed during the course of this work.The evaluation of compliance to the FAIR principles is one of the key issues we are facing. The establishment of the evaluation criteria reveals that compliance to the FAIR principles cannot be seen as an all-or-none situation but rather as a FAIR compliance spectrum ranging from not FAIR to FAIR and FAIRer. One of the challenging aspects for evaluating FAIRness is the assessment of the metadata. Indeed, rich metadata description of data is at the core of the FAIR principles. However, nothing actually establishes what rich metadata means for a data artefact or a semantic artefact. In this presentation, we will describe our attempt to establish such a threshold to FAIRness for semantic artefacts by defining a minimum metadata schema for them to be FAIR. This work stems from the collaboration of the FAIRsFAIR project and the RDA Vocabulary and Semantic Service Interest Group. During this work, we investigated the creation of a minimum metadata profile for Semantic artefact based on W3C DCAT 2.0 and the Metadata for Ontology Description and Publication Ontology (MOD) (Dutta et al., 2017). The MOD 1.4 ontology has been designed by reviewing and integrating existing metadata vocabularies (e.g., Dublin Core, OMV, VoID,...). More than a hundred metadata elements have been selected and integrated into the MOD ontology. This work contributed to the extension of the MOD ontology (MOD 2.0) with additional classes and properties derived from the DCAT model (subclasses) and specific to semantic artefacts e.g., Semantic artefact (subclass of Dataset), Semantic Artefact Catalog (subclass of Data Catalog), ... In addition, a list of metadata fields from MOD has been identified to be part of the minimum metadata model. To establish a first consensus on this minimum metadata schema, a workshop has been organised in 2021. This workshop involved a large number of semantic and metadata experts from various disciplines and was used to introduce the model and to decide through a voting process which of the DCAT and MOD fields should be mandatory, recommended or optional. The results of these votes are described in Le Franc et al., 2022. Based on the results of the vote, we created an OWL and a SHACL representation of the minimum metadata schema to support the publication of semantic artefact metadata and used in a reference implementation of the FAIR Data Point (FDP). To test this metadata schema, we leveraged the federated FAIR Data Space technology (F2DS) developed within the EOSC Pillar project (Cazenave et al., 2020) to align the metadata content of three different semantic artefact repositories (Bioportal, Agroportal and the NERC Vocabulary service) to the newly generated Semantic DCAT Profile. The converted metadata is published into a FDP instance which allows browsing the metadata through the FDP client User Interface and querying using the FDP SPARQL endpoint and/or RESTful API. This FDP instance is a first prototype for a semantic artefact search engine offering the possibility to search for cross-disciplinary semantic artefacts as proposed in Goldfarb et al. (Goldfarb et al., 2017). This first minimum metadata schema is proposed in machine actionable formats which can then be easily integrated by FAIRness evaluators for Semantic artefacts such as FOOPs and O’FAIRe. This work should be continued in the context of the RDA VSSIG and the Horizon Europe project FAIR Impact.