Proposal: Add a PMD publication metadata profile for citable semantic assets

[mbruns91]

Motivation

I am currently working on metadata extraction from PMD/PMDco-based RDF artifacts for publication workflows, especially for Dataportal and Zenodo publication via praeco.

The immediate use case is: given a semantic asset, e.g. an instantiated rdflib.Graph or a Turtle serialization, software should be able to harvest as much publication metadata as possible automatically:

• title
• description
• creators/contributors
• license/rights
• DOI or other persistent identifiers
• version
• language
• related identifiers
• resource/distribution links
• provenance/source information

PMDco already provides a strong MSE domain model, especially around materials, processes, data items, devices, qualities, and information content entities. However, for publication tooling it is still difficult to know which node in a graph is the citable publication unit and which triples should be interpreted as publication metadata rather than domain facts.

Proposal

I would like to propose adding a small, explicit PMD publication metadata profile rather than expanding PMDco with many new publication-specific classes.

The goal would be to define a recommended pattern for citable semantic assets that reuses existing standards wherever possible, for example:

• DC / DCTERMS
• DCAT
• DataCite-style identifiers
• schema.org
• PROV-O where applicable
• BFO/RO/IAO-compatible PMDco v3 modelling patterns

Concretely, the profile could define or document:

1. How to identify the citable artifact root

   • e.g. the RDF graph, semantic dataset, publication package, file data item, or distribution that should be harvested for publication metadata.

2. Which predicates should be used for bibliographic metadata

   • title
   • description
   • creators
   • contributors
   • publisher/contact
   • license/rights
   • version
   • citation
   • DOI and other identifiers
   • related identifiers

3. How to distinguish scientific/domain roles from publication roles

   • e.g. operator, instrument owner, project participant, data curator, publication creator, publisher, funder, rights holder.

4. How to model resources and distributions

   • RDF graph
   • original CSV/data files
   • generated metadata files
   • external resource URLs
   • file formats and media types

5. How provenance should connect to publication metadata in PMDco v3

   • especially since PMDco 2.x used PROV-O patterns while PMDco 3.x is BFO-based.

6. A SHACL shape or validation profile

   • A practical SHACL shape would be especially helpful for tooling.
   • Required/recommended fields could be validated before upload to Dataportal, Zenodo, MatPortal, or other repositories.

Why this would help

This would make PMDco-based RDF artifacts much easier to publish and reuse.

It would also reduce the amount of project-specific heuristic code needed in downstream tools. For example, praeco could implement deterministic metadata harvesting instead of guessing whether a dc:title belongs to the citable dataset, a CSVW column, an ontology, or a domain object.

This would also support better interoperability with Dataportal, Zenodo, MatPortal, and other RDM/catalog systems.

Suggested direction

My preference would be:

• keep PMDco focused on MSE semantics,
• reuse established publication/catalog vocabularies,
• add a small PMD-specific profile/pattern for connecting the MSE graph to publication metadata,
• provide a SHACL shape and example graph.

Offer to contribute

I am happy to help implement this to the extent of my abilities in semantics.

In particular, I could contribute:

• an initial example Turtle pattern based on existing PMDco 2.x / PMDco 3.x data,
• a first SHACL shape proposal,
• documentation/examples for publication metadata harvesting,
• follow-up PRs once the preferred modelling direction is agreed on.

My immediate downstream use case is implementing RDF metadata harvesting in praeco, but I think the same profile would be useful more broadly for PMD publication workflows.