Aligning with the FAIR Principles is a key requirement for European-funded research. However, diferent interpretations of the FAIR Principles lead to diverging evaluations of how to implement them. Previous research introduced the FAIR Implementation Profile (FIP) as a driver to accelerate broad community convergence on FAIR implementation options. To scale FIP creation and analysis, we established a decentralised socio-technical ecosystem, supported by a lightweight FIP ontology including a typology of FAIR Supporting Resources (FSRs). However, categorising FSR instances is sometimes challenging, suggesting that there are unclarities in the FSR type definitions. This paper presents an ontological analysis of the FSR typology, aiming to improve its accuracy in supporting the peer-reviewed curation process of FSR descriptions provided by a wide community of FIP users. Using the Unified Foundational Ontology (UFO) as a reference, we analysed class definitions and demonstrated UFO's capability to resolve controversies leading to disambiguation of FSRs of type FAIR Specification and FAIR Practice.
The FAIR Principles [
To make FAIRification more systematic and eficient, the GO FAIR Foundation (GFF), in
collaboration with the ENVRI1 initiative, developed a questionnaire-based approach to capture
community-specific FAIR Implementation Profiles (FIPs) [
These developments require up-scaling approaches to FIP creation, dissemination and analysis.
In recent years, a decentralised, socio-technical ecosystem supporting the composition of FIPs,
their representation, curation (formal peer-review) and discovery has been developed. The
core of this ecosystem is a lightweight FIP ontology3 including a typology of FAIR Supporting
Resources (FSRs). FSRs are resources that support the FAIRification process. FSR subclasses
represent resource types that are essential to implement the FAIR Principles, the so-called FAIR
Enabling Resources (FERs), instances of which are used to compose a particular FIP. Creating a FIP
is not a trivial task and it is advisable to get support from a FAIR expert. To guide communities in
going FAIR, GFF has developed a Three Point FAIRification Framework (3PFF) with an associated
Capacity Building Programme to train data stewards on how to facilitate FIP events with the
participation of interested communities of practice [
Therefore, to improve the common understanding of FIP-related concepts, this paper aims to start an ontological analysis of the FIP ontology, focusing on resolving classification issues within some FSR types. By refining these definitions, the FIP ontology should be better suited to serve as a semantic layer for providing a reference knowledge base FSRs. Furthermore, we tried to identify important properties to be defined in the metadata schemas for describing the instances of diferent types. However, a key challenge is to maintain backward compatibility within the system, to ensure that legacy FIPs and existing users can continue using it without disruption.
The paper is further organised as follows: Section 2 provides background information on related work for landscaping the communities’ infrastructure and for ontological analysis aimed to clarify its representation; Section 3 describes the components of the FIP ecosystem; Section 4 explains the FIP ontology; Section 5 analyses detected curation challenges of communityprovided FSR instances; Section 6 exemplifies, using the FSR description of the Research Object Crates, how ontological analysis can help reveal ontological shortcomings and how to resolve them; Section 7 concludes with a future outlook and next steps.
The FIP approach is accompanied by the definition of resource classes explicitly designed to
address each of the FAIR Principles. This was preceded by a thorough analysis of the FAIR
Principles by considering how various aspirational FAIR behaviours might be instantiated
in diferent types of (mostly existing) resources. The type definitions of the FERs follow the
interpretations published on the GFF website4, which are based on an analysis made with the
involvement of a broader expert community [
By providing FIPs, communities contribute to a knowledge base of FSRs, which can be found
and reused by other interested parties. To be findable, the resources must be described following
an agreed pattern and classification. Note that repositories for collecting community standards
and technologies are being ofered by FAIRSharing 7 [
While the value of FIPs for facilitating the FAIR onboarding of communities has been
repeatably demonstrated [
To support the creation of FIPs, the “FIP questionnaire”9 was compiled, which is a set of questions
covering the FAIR Principles, with the answers identifying resources enabling the FAIRification
of (meta)data. FAIR answers to these questions are given by means of a FIP ontology, which
semantically defines the diferent types of resources empowering the FAIRification work. Initially
it described resources essential in enabling the FAIR Principles (FERs), but it was expanded to
include all resources supporting the FAIRification process, using the broader FSR type. Using
a similarly conceived “SIP questionnaire", these resources can be compiled into a Semantic
Implementation Profile (SIP) . We developed the following collaboration tools to leverage the FIP
ontology (see Figure 1):
• Nanopublications [
publication mechanism. • The survey wizard hosts, like the FIP12 and SIP13 Wizard, are two Data Stewardship Wizard (DSW) instances, that make use of knowledge models to design project templates. In our context, the knowledge models are based on the FIP ontology to create questionnaires that allow the compilation of FIPs and SIPs respectively, and their publication in diferent formats, including nanopublications. • The FSR curation GitHub repository14 provides an environment for curating FSR descriptions. A Github action15 automatically lists new FSR nanopublications, which are then reviewed by at least two GFF trained FAIR experts, following well-documented curation guidelines and a process of iterative improvements. The curation process assesses the usefulness of a resource in its function to FAIRify data, and verifies its correct alignment with the appropriate FSR classes based on their definition in the FIP ontology, but does not assess the degree of FAIRness. It also provides a quality check of the completeness of the metadata description and suggests improvements where appropriate. A successful review leads to an approval nanopublication (using a dedicated template16) and a curation badge being assigned in the wizard. This process helps prevent duplication, inadequate descriptions, unhelpful FSRs, and maintains the integrity of the FIP approach. In addition, the process is documented, providing an invaluable insight into the debates and diverging interpretations leading to the final FSR classifications. • FAIR Connect17 acts as an index and search engine for finding published FSRs. The search results include metadata information, such as the GFF curation status and popularity with FAIR Implementation Communities (FICs).
The FIP ontology, as shown in Figure 2, was conceptualised in 2020, introducing five main classes and associated properties in the FIP ontology. At the core of the ontology is the FIP Declaration class that states that a FIC declares the use of a FER. The FIP Declaration refers to a FIP Question that refers to a FAIR Principle. FERs are essential to the operationalisation of the FAIR Principles, providing the functions needed to achieve some aspect of FAIR behaviour. By request of the ENVRI communities, we introduced a more flexible representation of the FERs, allowing the description of available FER and FER in development as sub-classes of FER to reflect the rapid developments in the area of FAIR implementation. In alignment with these two sub-classes, we introduced flexible properties to define the use of these FERs: next to declares-current-useof, which must be an available FER, one can declare-planned-use-of a FER, which should be accompanied by a declares-planned-replacement-of a FER. Each FIP Declaration is complemented 12https://fip-wizard.ds-wizard.org/ 13https://sip-wizard.ds-wizard.org/ 14https://github.com/gofair-foundation/fsr_curation 15https://github.com/gofair-foundation/qualification-issue-creation-action 16https://nanodash.petapico.org/publish?25&template=https://w3id.org/np/RAi7_
UxEF3TTPBp7lmWOVvKR-jUmZgDfY_ZIe57dqbOnQ&template-version=latest 17https://fairconnect.pro/ by the data type property considerations to allow the description of the community requirements and constraints that led to the implementation choice.
Figure 3 gives an overview of the types of FERs used to address each of the principles, which can be grouped into three main types: services, specifications and policies. However, in practice, an FER instance can encompass multiple subtypes within one parent type, like DCAT18, which is a specification for a Metadata Schema, Structured Vocabulary, a Semantic Model and a Provenance Model.
To provide more information about instances of the diferent FSR types, metadata templates in nanodash and corresponding metadata description templates in the DSW are provided. We used core metadata elements for all types like rdfs:label, rdfs:comment, rdf:type, and rdfs:seeAlso, plus specific elements for each of the types to better describe their particular features. For example, in the nanopub template for identifier services , we added the predicate implements using the object property URI http://usefulinc.com/ns/doap#implements to define which specification the service is based on. In this way, the templates enable relationships between FSR nanopublication instances.
The EOSC Association’s Semantic Interoperability Task Force designed a survey to capture the
semantic interoperability practices from across diferent communities of practices, consequently
the FIP ontology was extended with additional concepts and relationships to address semantic
interoperability aspects [
The SIP application of the ontology adapted the FIP Declaration concept to SIP Declaration, by introducing two new optional relations with the associated objects: declared-for-case-study with range: FAIR case study and declared-for-digital-object-type with range: Digital Object Type. These relations were then also taken up in the FIP approach, as providing explicit choices for a specific digital object type makes the profile more meaningful. In terms of the FSR types included, the SIP focuses on Metadata Schemas, semantic artefacts (which include Structured Vocabularies and Semantic Models), Provenance Models, but also on Crosswalks (e.g., mappings between schemas). In addition, it uses services for the generation, transformation, and validation of resources, such as Editor, Validator Service, FAIR Representation Service, Web API, provenance Tracking Service. FAIR Practice was introduced as a class for representing combinations of specifications and services. These new classes required an extension of the FIP ontology beyond the FER typology, introducing the FSR as a superclass with FAIR Supporting Services and FAIR Specifications as main subclasses (see Figure 3). For completeness, it should be noted that the FSR typology includes other classes, such as, e.g., FAIR Data Stewardship Event, which are ignored here since they are beyond the scope of this paper.
Considering that the FIP is a collection of community-selected (or -created) resources that provide the answers to the various FAIR Principles, and an FSR can be any resource created in any context, it is understandable that achieving a straightforward one-to-one correspondence between principles and resources is more of an exception than a norm. This is not inherently problematic (after all, FSRs can provide -sometimes partial- solutions to multiple FAIR concerns), but it does lead to ambiguities, for instance when FSR types with potentially competing scopes are combined. The assignment of FSR types is currently subjective and disagreements can occur, which becomes evident in the peer-review curation process documented as GitHub issues.
An FSR that showcases this phenomenon is Research Object Crate (RO-Crate)19, as reported 19https://www.researchobject.org/ro-crate/ in GitHub issue 11920. A nanopublication for this FSR is available21, and the GitHub issue also points to an existing duplicate which has only Metadata Schema listed as its FSR type. Both nanopublications were authored by the same individual, so the nanopublication with the older date should have been replaced with the newest one. The latter describes RO-Crate as a Metadata Schema, Structured Vocabulary and Semantic Model, which is entirely justified, because, as pointed out in the documentation22, the metadata schema is defined in JSON-LD, reusing schema.org vocabularies and includes qualified relations to other standards like Portland Common Data Model23 and Dublin Core Terms24. As argued by the lead of this paper (GitHub user mabablue), additional types should be included in the description: 1. Provenance Model ,as also described in the RO-Crate project in the FIP Wizard25 created (but not yet available as a nanopublication) by Stian Soiland-Reyes26, who is a co-developer of RO-Crate; 2. Metadata-Data Linking Schema, as RO-Crate packages research data with their metadata.
When all categories apply, the categorisation might lose its usefulness. On the other hand, specifications can include all these diferent aspects, as it applies in the example given. However, when constructing a specification type FSR, it is best practice not to mix FAIR Supporting Services and FAIR Specification . A potential alternative is to consider the RO-Crate specification to be a complex FSR, composed of multiple sub-FSRs with their own types. Another option would be to categorise the RO-Crate approach as a FAIR Practice, which suggests the combined use of a set of FSRs to reach a FAIRification goal.
The FIP ontology started with twelve FER types focusing on the implementation of the FAIR
Principles. However, it gradually expanded over time, driven by requests of the communities
of practice, such as those from the ENVRI-FAIR project. This expansion requires refinement
to provide clear descriptions of FSRs, and to ensure interoperability with existing relevant
semantic artefacts. Not refining the ontology bears the risk of inconsistencies, inaccuracies
and/or bad practices creeping into its application for FSR descriptions. To mitigate this risk,
we performed an ontological analysis focusing on some aspects of the FSR typology, based on
the Unified Foundational Ontology (UFO) [
To facilitate the readability of the notions used, we apply diferent styles from here on:
UFO concepts and «meta-properties». The parts of UFO [
UFO makes a fundamental distinction between Individuals, which are entities existing
in reality with a unique identity, and Universals, which are abstract patterns of features
that can manifest in various individuals. Unlike Perdurants, Endurants are individuals that
are wholly present whenever they are present. Substantials are existentially-independent
Endurants, whereas Moments are individuals dependent on their bearers. Substantial
Universals are kinds of universals whose individuals are substantial individuals [
UFO-C [
Mapping the UFO concepts and meta-properties to the FIP ontology it becomes clear that all FSR instances can be considered as Individuals while FSR types can be considered as Universals. Further, we can map FSRs to Endurants and «Categories», as they represent an aggregation of diferent «Kinds». As defined in the FIP ontology, FAIR Specifications are “precise descriptions of features, requirements, constraints” and therefore we suggest to map them to Normative Descriptions, which are endorsed by one or more FICs. FICs can be mapped to Collective Social Agents that bear Intentions to implement the FAIR Principles. All specific FAIR Specification types like Metadata Schema can be considered to be «Subkinds». FAIR Practices can be interpreted as Plan Descriptions (see Figure 5) that describe Complex Actions like the proper usage of a required FSR or a combination of FSRs to achieve a FAIRification goal. FAIR Policies can be distinguished from FAIR Practices by the level of commitment of FIC members, to align with the rules defined in the policy. FIP Declarations are Perdurants that might be classified as Communicative Act that can result in Normative Descriptions, like the FIPs which are recognized by FICs.
Considering the possible classification options for the FSR instance RO-Crate, it can be inferred from the mapping provided that: • Complex FSR: a specification may intertwine diferent specification types like Metadata Schema and Provenance Model in one semantic artefact, and thus may not be individually divisible.
• FAIR Practice: is a description of how to carry out a complex action that requires FSRs,
which may include specifications. • Multiple subtypes of FAIR Specifications may be applied, but it is crucial to avoid mixing between «Kinds» (services and specifications). This rule is not reflected in the ontology and therefore we recommend to explicitly state this rule in the guidelines for the FSR instance descriptions.
The analysis results in a refined definition for FAIR Practice: “A method description detailing the use of a specific resource or a combination of resources as applied by a community to achieve FAIR processing of information.” Additionally, the metadata template for this type has been enhanced as follows: • To include statements about employed FSRs the
http://purl.org/dc/terms/requires is used. • To include statements about the supporting FIC the https://w3id.org/fair/fip/terms/declared-by is used. object
predicate object property We applied the improved template to describe the RELIANCE Research Object Practice, which recommends the use of the FAIR Specification RO-Crate and the Registry ROHub27, and the GFF OSF Practice28, which advocates to register all the relevant artefacts created for a 3PFF event in the Open Science Framework. We will also provide clear guidelines on how to provide FSR instance descriptions to prevent poor type allocations. We plan to verify whether the suggested changes reduce ambiguities regarding the classification of FSRs with multiple types by analysing issue resolution in the Curation GitHub repository. 27https://w3id.org/np/RAq2HHTwWLjZzP-gIJIRJiQJpjd4KTrMn-WLvnKo1mXdU 28https://w3id.org/np/RADBsQ41yWver59fedlE0kx3gUDaWMplqM8-BzgOYxUz8
By conducting ontological analysis using an upper ontology such as UFO as a reference, it becomes possible to ofer clarifications in the classification process when submitting a nanopublication description of a specific FSR. This is also instrumental for qualifying FSR descriptions. As demonstrated in this paper on the FAIR Practice example these clarifications will help to improve the concept definitions, the user guidance text in the wizards and the metadata templates for FSR instances. We could also clarify that RO-CRATE is best described by classifying it to multiple FER types of kind specification. Further analysis is needed to enhance the FIP ontology and ifne-tune the definitions of its classes for other FSR types. Moreover, we want to improve the metadata descriptions of FSR types by analysing relations between diferent types. Validation of the improved ontology and metadata schemas will be performed using real-world example FSRs, covering both new instances and existing resources to ensure backwards compatibility.