Currently the specification has definitions for classes, but not for all of the properties used in the metamodel. Some definitions are available in the ontology for these properties, but the metamodel section should be clarified to include them.

While typically one would associate a vocabulary entry with only one MVF entry, this does not allow for cases when a vocabulary includes more entries than are mapped to MVF entries. Thus, one could have terms that either are not needed or have yet to be mapped, and the restriction would result in wrong results (or a rule that exercises the restriction could do so).

The restriction in the MVF ontology should be modified to 'max 1' from exactly 1, and the redundant restriction in the ISO 1087 ontology that says 'min 0' should be eliminated.

MVF element is the parent class of both MVF entry and vocabulary entry. A given term, which is a vocabulary entry may be used in various ways, and in some cases may have multiple textual names. This isn't common, but does happen in practice, especially in cases involving legacy data.

One case we have found on the IDMP-O project is that there are two sources for the same term. The European Medicines Agency publishes controlled vocabularies are specified in the EMA SPOR reference ontology (RMS). That ontology includes not only EMA SPOR controlled vocabularies but duplicates a subset of the controlled vocabularies from MEDdra (Medical Dictionary For Regulatory Activities). In the IDMP representation of various pharmaceutical products, both controlled vocabularies are required for use in the EU. Thus we have the EMA SPOR version of the MEDdra term and the actual MEDdra term, which we map to one another using owl:sameAs. That results in a logical inconsistency because we have more than one term name for the same term. We have similar issues with certain FDA controlled vocabularies, where they have identifiers specified in the GSRS and UNII repositories, that are actually the same identifier, thus resulting in two textual names for the same thing.

There may be a requirement for some implementations to use a SHACL shape for this, but for information content, such as the specifications for certain substance names, simply relaxing the restriction from exactly one to some values from would provide the semantics we need.

A numeric ordering is a kind of systematic ordering, with a couple of important attributes: (1) what the origin is – zero or one, typically, and (2) whether it is from low to high or high to low. Typically numeric orderings of collections use positive integers, but there could be other kinds of ordering that use, for example, alphabetical with numerals and/or distinctions by rank, such as A.1.1.

There may also be utility in having the notion of 'position', such as a position in an ordered list, or position of a term in a text.

We have not captured this clearly in the ontology and should.

The label in the terminology science ontology reads 'term harmonization' and should be 'term formation'.

This issue affects only that ontology, not the documentation for it.

Currently we've mapped MVFEntry to SKOS Concept, and it should be Vocabulary Entry to SKOS Concept. Likewise, the mapping between MVF Concept System to ConceptSystem should be to Vocabulary in MVF.

Investigate whether this should be on vocabulary, vocabulary entry, both, and/or possibly on a more general element as well.

There may also be an impact on the ontology, depending on the resolution.

This represents a generalization as in UML or OWL, at the metamodel level rather than at the UML model level. This may be confusing to users and so an explanatory note is needed to help. Note that the ramifications of this are at the instance level - one instance of an MVF entry may be a generalization of another instance of an MVF entry.

The same is true of the additional relationship, which is context, where notes and examples would also be helpful.

While the definition of MVF entry and that of SKOS concept appear to be the same, there are differences in how SKOS is used. That may, in fact mean we should consider mapping to MVF Vocabulary Entry instead.

Regardless, it would be useful to provide a discussion in the specification about the possibility, depending on the application.

The property 'currentMVFEntry' on the association ElementCurrentEntry changes the multiplicity from 0..* to 0..1, so it should be redefines not subsets (or the parent multiplicity needs to be changed)

This might be a new area of exploration, and could apply beyond MVF entries in the context of MVF.

Currently the metamodel reflects SKOS-like relations such as broader and narrower, but more is needed for certain applications.

It would be useful to have a non-normative annex demonstrating how one could do this as a starting point for MVF users.

Just prior to publication of the FTF report, a new property was added to the metamodel called descriptiveReference. This property, which is quite useful, was not added to the ontology, however, and may eliminate the need for restrictions using the Commons is defined by (cmns-dsg;isDefinedBy) property on VocabularyEntry.

MVF users have requested that the form of the XMI that is provided as a part of the MVF specification be canonical XMI.

There is an example in Annex B that is quite useful, but is not multilingual. Since we have said that MVF can be used to support multiple natural languages, we need at least one additional example to cover that use case. We might also include a broader example showing how to represent a general vocabulary in MVF, such as the AI PTF emerging vocabulary for artificial intelligence, to show how this can be done.