By José Emilio https://orcid.org/0000-0001-8907-5348

Review Details

---

Reviewer has chosen not to be Anonymous

Summary of paper in a few sentences:

The paper describes a Knowledge Graph Metadata Specification, a set of 33 metadata elements that can be used to describe linked datasets. The paper describes the methodology used to obtain those properties and have also created a github repository to maintain that vocabulary. The paper also presents how those terms can be used to describe one Knowledge Graph.

Reasons to accept:

The paper is very well written and is easy to follow. It describes the process to identify those metadata elements and provides a comparison with previous approaches and a well structured discussion about the status and the future of the work.

The paper also has some supplementary contents in 2 github repositories, one for the Knowledge Graph Metadata Specification and another for the Food Claims Ontology.

About the metadata itself, I think the set of properties is good enough as it has been proposed by a community of experts. I wonder if there should be some mechanism established to maintain the set of properties in a more collaborative way…maybe having those shapes maintained in a spreadsheet like format from which they could be exported to SHACL (or ShEx). I may suggest the use of DCTAP for such templates as it has been done recently by PubChem: https://pubchem.ncbi.nlm.nih.gov/docs/rdf-schema where they keep the source of truth in a DCTAP spreadsheet and generate ShEx, SHACL, etc. shapes from it.

In fact the authors chose to publish and maintain the shapes in SHACL, which can be OK, but looking at the shapes, I think the shapes could have also been published in ShEx and could also be more readable and accessible to other processors implemented in ShEx.

In the introduction, when the authors describe related works, I would suggest them to mention rdf_config: https://github.com/dbcls/rdf-config which is used by DBCLS and also provides several metadata possibilities with the inclusion of example SPARQL queries.

In the case of the KG Schema field with the property dct:conformsTo, I wonder if it could contain more structured values than just an IRI. The reason is that there are different technologies for validation like ShEx/SHACL/… and it may be necessary to indicate that it has to conform to some schema using some technology or even some format (Compact syntax vs RDF syntax), so an automatic process could validate the contents.

In the case of the version, the value is a literal. Would it be possible to extend it to use IRIs to identify versions?

Listing 1 has two values for example, sh:example and :example, but I think the SHACL vocabulary doesn’t have the property sh:example

I wonder if the authors have taken into account other properties to describe SPARQL endpoints like the propertyPartition, classPartition, etc. which could be useful to both human users and clients to optimize query plans based on those numbers.

Reasons to reject:

I found no reason to reject the paper. Maybe the paper is not a full research paper providing something new but it can be a useful resource.

Nanopublication comments:

Further comments:

Maybe, the authos should try to promote their work so it can be adopted by the community.

Reviewer has chosen not to be Anonymous

Summary of paper in a few sentences:

The paper proposes a comprehensive metadata specification and SHACL-based validation framework for Knowledge Graphs, developed through a community-driven process and comprising 33 metadata elements. The authors map the specification to the Linked Open Data (LOD) Cloud and evaluate 1,573 datasets, revealing widespread inconsistencies and missing metadata. They also demonstrate the framework by publishing FAIR-compliant metadata for the Food Health Claims KG, showing improved machine-readability and discoverability.

Reasons to accept:

- The paper adopts a rigorous, community-driven methodology for defining the metadata specification. By involving domain experts and iteratively refining the specification through workshops, the authors ensure that the resulting framework reflects shared needs and established best practices within the Knowledge Graph community.
- The paper addresses a well-recognised challenge: the lack of consistent, high-quality, and FAIR-aligned metadata for Knowledge Graphs. This problem is clearly articulated and highly relevant, as it directly affects KG discoverability, reuse, and interoperability—core concerns for researchers and practitioners.
- If widely adopted, the proposed specification and validation framework could significantly improve metadata quality and standardisation across KG ecosystems. Its alignment with existing vocabularies and FAIR principles positions it as a potentially influential contribution that could support improved KG publishing practices and enhance metadata-driven discovery mechanisms.

Community-driven development: The specification is the result of a collaborative expert process, supporting its legitimacy and potential uptake within the KG ecosystem.

Reasons to reject:

- The paper does not convincingly demonstrate why established vocabularies such as DCAT, DCTerms, PROV, VoID, or Schema.org are insufficient for describing Knowledge Graph metadata. The specific gaps or shortcomings in these models are not analysed in depth, leaving the motivation for a new specification underdeveloped.
- The proposed framework merges terms from multiple vocabularies with the goal of enhancing interoperability. However, combining heterogeneous vocabularies without a unified semantic foundation risks introducing inconsistencies, ambiguous term reuse, and misalignment with the original conceptual intent. The paper does not explain why defining a new vocabulary with its own namespace and semantics, later aligned to existing standards, would not provide a clearer and more robust solution.
- While the validation of 1,573 LOD Cloud datasets identifies substantial metadata gaps, the analysis remains mostly descriptive. The authors do not examine why these deficiencies arise—whether due to limitations of the LOD schema, community practices, or tooling—and do not convincingly demonstrate how their proposed framework would materially improve metadata quality or adoption at scale.
- The paper does not discuss established efforts such as the EU Metadata Quality Assurance framework of the European Data Portal, which offers automated checks on metadata completeness and consistency. This omission weakens the contextualisation of the work within the broader ecosystem of metadata quality tools and standards.
- The demonstration is limited to a single Knowledge Graph produced by the paper’s contributors. This restricts the generalisability of the findings and may raise concerns regarding objectivity, as the framework is not tested on independently developed KGs or in contexts where the authors have no direct involvement.

Nanopublication comments:

Further comments: