PREreview of DRAFT: Developing and implementing the semantic interoperability recommendations of the EOSC Interoperability Framework

Those are my comments organized following the 5 Recommendations. I may sometime include there comments related to the rest of the report, but went in details on the Recommandations text mostly.

1. Align emerging adaptations and implementations of the EOSC - IF reference architecture

This recommendation is a call for a commitment to the EOSC-IF which I see very positively.

In a sense it also witnesses/recognizes the main limitation of the EOSC-IF: the fact that it stays very theoretical and does not rely or build on some clearly identified –and available– technologies. In other words, Recommendation #1 says “leave multiple possible implementations of the IF possible and align them” and also recognize these implementations/adaptations are “emerging”. If those are emerging, maybe a stronger requirement would have been to rely from scratch on some technical commitments rather than watching the diverging implementations emerge and then work on their alignment. Although, I recognize the statement of the task force is probably more realistic than mine here. 

I would have phrased the 4th indicator (“shared vocabulary”) more explicitly and used the plural form. This sentence seems to suppose there will be “one shared vocabulary” whereas it has been historically demonstrated by philosophers aside/within the semantic web community that a unique ontology or vocabulary to “rule them all” is impossible to build. It never converges. The only approach is to work with a diversity of ontologies/vocabularies properly served and aligned one another. This diversity approach is in fact the one encouraged by Recommendations #4 and #5. So I see a bit of a contradiction with the “one shared vocabulary”. 

In the 4th suggested action, I think the expression “semantic view of the EOSC-IF” is ambiguous. Idem, in the 5th suggested action, I do not really understand the expression “number of terms related to semantic interoperability”. Especially with the use of the word “terms” that is often used to identify objects (classes, concepts) in semantic artefacts.

2. Identify and consolidate different approaches to representing and exchanging (meta)data with the FDO model described in the EOSC - IF

I am not a fan of the commitment to the FDO Framework. Simply because I have not seen this framework implemented in practice and I am not sure it will ever be. Still the vision of the FDO Framework of a new protocol extending HTTP is very interesting but I am not sure it is realistic. HTTP is the protocol on which all the Web works and touching it is very very much complicated (see https://www.w3.org/Protocols/). Attaching the destiny of “interoperability and FAIR” within EOSC to the destiny of the FDO Framework to me seems a risk that I am not sure we should take. I mean, we can build a “Web of FAIR data and services” with the currently available technologies on which the Web is based, and especially the Semantic Web technologies standardized by the W3C. 

Also, in the Recommendation text, I would not use the expression “FDO model” which is not the one used by the FDO developers. They talk about the “FDO  Framework (https://fairdigitalobjectframework.org/ and https://www.go-fair.org/today/fair-digital-framework/). The expression “FAIR digital object model" has only 278 hits on Google (one from me, my mistake too). The expression “FAIR digital object framework" has 914 hits including the ones from the reference web page.

Phrasing of the 1st indicator is ambiguous: “how they define '' we do not know which word the pronoun “they” refer to: the community? guidelines? availabilities”? Also the expression “boundaries around (meta)data” is very ambiguous. Suggested phrasing: “"More community guidelines related to semantic interoperability and (meta)data and the use of semantic artfacts."  

In the 2nd and 3rd indicators: the word “references” makes me think that the only/good way to go is to have (meta)data refer to semantic artefacts content. I mean the challenge is not to “refer” to standard vocabularies/terminologies or ontologies but really encode (meta-data with them i.e., represent the data natively using the terms provided by the vocabularies/terminologies and relying on the schemas defined by the ontologies. Not just refer to them. I would merge 2nd and 3rd indicators into one with the expression “qualified and resolvable references” then I would make another 3rd indicator such as:” (Meta)data is increasingly encoded or represented with concepts and classes defined in semantic artefacts or rely on their schemas.” This suggestion can also be used to rephrase the 2nd suggested action.  

I would complete the 3rd suggested actions with “and ease the interoperation of these editor services with relevant semantic artefact catalogues.” Indeed, this is crucial to see the lifecycle of semantic artefacts from a complete point of view from designing/building them (in “editor services”) to sharing and serving them (in semantic artefact catalogues). In concrete words: be sure tools like Protégé and VocBench work closely with repositories such as OntoPortal ones or OLS, NVS etc. 

3. Extend the EOSC - IF to include a research process perspective that can support convergence on solutions for common use cases

I agree with the recommendation to follow, get inspired and build from what is already done and observed in communities, called “use cased” in the text. 

First indicator is indeed very important. I would remove “and other actors” and “and other functional content” from the text as I believe they do not bring anything more to the sentence and even create ambiguities. It's like using “etc.” in a listing: sometimes it makes the point more fuzzy than expected.  

In the 2nd indicator the word “language” is a bit ambiguous.

In the 4th suggested action, I would add “and FAIR-IMPACT projects” has FAIR-IMPACT is in sense the continuation of FAIRsFAIR, especially on the question of semantics, in which the project goes further than FAIRsFAIR that has established the baseline discussion. FAIR-IMPACT’s WP4 is entirely focused on ontologies and metadata and has explicit tasks on multiple aspects of semantic artefacts (their governance, their FAIR design, their catalogues, their mappings, their use in data repositories).

4. Extend the set of semantic objects described in the EOSC-IF to include artefacts such as mappings and crosswalks

I very much like Recommendation #4 because of the importance of mappings and crosswalks in the EOSC context. I very much agree with all the indicators and suggested actions. Overall the Recommendation aligns very well with what is discussed/proposed in FAIR-IMPACT T4.4, which currently plays the role of a forum (with multiple public workshops and the creation of a new RDA WG) to discuss this matter. However, I would push the recommendations to explicitly recognize and recommend SSSOM (see previous comment on the fact that the EOSC-IF is too abstract in terms of technical commitments).

I believe the sentence of the Recommendation uses two words that are super abstract: objects and artefacts. I would turn the text in a way to use only one. Also, I think the objective is written in an ambiguous way: “reusing the related semantic objects to support mediation across semantic artefacts”. What do you mean exactly? Another reason to remove the expression “semantic objects” that I believe create an ambiguity with semantic artefacts. Are the objects the content (classes, concepts, properties, instances) of the semantic artefacts? 

The report has a full page on “Mapping repository” (page 16) which does talk about mappings and crosswalks, but not really about the fact of storing/hosting and serving these mappings in repositories or services. Maybe consider including in the report more elements related to the fact that ESOC needs to provide platforms and services to host, share and serve mappings (and maybe also ease their creation and curation). 

In the title of the section p. 16 and overall in the Recommendation text, I would not use the expression “The mapping repository”. It seems there will be one and only one mapping repository for EOSC (and more) which I think is not realistic and not desirable either. I would use the plural form “mapping repositories”.

In the 2nd, 3rd and 4th indicators I would remove the word “component”. The expression “mapping repository” is already a name. In the 3rd indicator, I have the same comment as above on the expression “shared vocabulary”.

Consider merging 4th and 5th indicators. 

In 2nd suggested actions, the final sentence creates an ambiguity: you generalize “mappings, crosswalks and other interoperability enabler semantic artefacts to FAIR” and then call for making “these artefacts accessible in catalogues of semantic artefacts”. Therefore, one does not understand anymore if you recommend to store mappings and crosswalks in mapping repositories or in semantic artefact catalogues. I personally think that semantic artefact catalogues must also include a mapping repository component in their platform (such as what is done for example by the OntoPortals instances or the NVS, but not all semantic artefact catalogues on the market). But, in the context of the task force report, I would stick to distinguishing the two services well.

5. Recognise the semantic artefact catalogue as a critical part of the long-term viability of any research data infrastructure

Overall this is, without surprise, my preferred recommendation. Indeed the role of semantic artefact catalogues has been identified by the ESOC-IF but except FAIR-IMPACT’s T4.2 and T4.5, we do not see many EOSC projects that are working to design such catalogues as a core components of EOSC and research data infrastructures to ease semantic interoperability. Glad to see the work of the task force encouraging for a broader use, development and support of such catalogues. 

Maybe complete the objective with “as well as their use to annotate and/or encode (meta)data.” to make explicit the fact that catalogues must play the role of archive for semantic artefacts, but also the role of platform of services to actually use them to encode/annotate data. For instance, OntoPortal instances offer an Annotator service that is used to identify ontology terms within text. This is a service offered by these catalogues to deal with data ; beyond the fact of simply hosting and archiving semantic artefacts. 

Maybe separate the 2nd sentence of the 1st indicator into another specific indicator. Maybe complete the first sentence with: “Multiple FAIRness assessment tools for semantic artefacts are in use within the catalogues.” Add “Semantically rich” to the beginning of the 2nd sentence. 

Similar to Recommendation #4, I would use a plural form for “semantic artefact catalogues” and remove “the” and “component” to avoid making think there will be one and only one catalogues and emphase/recognize the importance of multiple semantic artefact catalogues addressing multiple community needs.

In the 2nd (then 3rd) indicator, complete the sentence with : “Plus, semantic artefact catalogues also play the role of long term archive for semantic artefacts.” 

Complete last indicator with: “toward being themselves trustworthy as any other data repository.” to emphasis the fact that as any other data repository, eventually semantic artefact catalogues shall also be considered trustworthy too e.g., get certification like CoreTrustSeal. 

In the 1st suggested action, I would remove “Develop” at the beginning of the sentence as there is now a paper being published produced by the task force (and including me as co-author but not in the task force) related to such a maturity model. 

In the 2nd suggested action “to address found” seems to be an English problem. 

Maybe consider: 

• Encourage communities to adopt, improve and build on semantic artefact governance models such as the ones reviewed by FAIR-IMPACT.

and 

• Encourage the federation of multiple domain specific or project specific semantic artefact catalogues.

as 2 additionals suggested actions. 

Maybe complete suggested action 4th with: “and encourage their role in the (community-specific) governance of semantic artefacts.”

Between Recommendation #4 and #5, I really like the emphasis on mappings and semantic artefacts catalogues. Only regret that the convergence between these two dimensions is not done. I strongly believe that one cannot design an ontology repository without providing some kind of mapping repository or mappings services; and the other way around: designing a mapping repository without relying on an ontology repository seems impossible. Therefore, I would have recommended technology providers to work on overall solutions that offer an ontology repository and mapping repository in the same environment/platform. This is for instance what OntoPortal or NVS is doing. 

Competing interests

I am not in the task force. However, I have participated to the very end of the work of the Theme 2 (on semantic artefact catalogues) to publish the maturity model cited in the report (https://arxiv.org/abs/2305.06746).