'has identifier for [...]' RDA/RDF properties

[briesenberg07]

Template testing group discussed the '(has) identifier for work' property template at 2023-08-08 meeting* (see 2023-08-08).

Property template is currently set to lookup an RDA Nomen resource in Sinopia. Catalogers would be expected to create a Nomen entity using the UWSINOPIA:WAU:rdaNomen:identifier:CAMS resource template, record a nomen string, scheme of nomen, and possibly other information about the Nomen, then return to the work or other previous resource template and lookup the new entity, entering it as value.

Today we discussed other cases in which instead of creating a Nomen resource, a dereferenceable IRI would be entered directly as the value for 'has identifier for work'. There was some support for this, in the event that a suitable resource (IRI) is available**.

BUT

@pan-zhuo reminded me of the guidance for using the IRI recording method with 'has identifier for work' (emphasis added to the portion which I find confounding)...

Recording an IRI
Record an IRI for an instance of a Nomen as a real-world object.
Do not record an IRI for the entity that is being described.

...and also of the fact that the range of this (object) property is rdac:Nomen. He says:

I wonder if we will ever find "actionable IRIs" in the wild that could be construed as RDA Nomens. A Wikidata IRI stands for a Work, not a Nomen.

For me this muddies the water considerably...as in, walking down a hall of mirrors, muddy...

Setting aside the question of whether IRIs from external VES which conform to the modeling of RDA resource entities can be found**, and assuming for the sake of this discussion that we have located IRI https://rda_work_001...

<https://rda_work_001> a rdac:Work;
   rdaw:10329 "The Story of RDA/RDF at UW"@en .

❓ Would providing this IRI as the value of rdaw:P10002 result in well-formed RDA??

* UW NetID access only
** Identifying semantic web resources described using non-RDA ontologies which adequately meet semantic requirements for use as an RDA entity is an ongoing discussion ...

[briesenberg07]

@pan-zhuo recommends guidance > Recording methods > Recording an IRI. This has:

An IRI of an RDA entity is not treated as a kind of Nomen because the value of a nomen string is an RDF literal.

If we are using canonical properties, we are under no obligation to meet rdfs:range requirements...?

An IRI can be recorded as an identifier if it is "stringified." This process transforms the value of the IRI into an identifier nomen string that cannot be guaranteed to be unique outside of a local application.

So, instead of recording an IRI as the object of a triple with property like 'has identifier for work', record an IRI for a Nomen resource which has 'stringified' IRI as nomen string value? ...

[briesenberg07]

Follow-up with resource-template testers

@AdamSchiff @CECSpecialistI @gerontakos @cspayne @junghaelee @JianPLee

I believe I was incorrect in our discussion on 'has identifier for work' at the 2023-08-08 meeting. At that time, we discussed providing values for 'has identifier for work' along the lines following. (For the sake of this discussion I'll assume that Wikidata items which are instances of literary work are semantically equivalent to instances of RDA Work, which in fact seems not to be the case... [another discussion].)

<http://example.com/MobyDick> a rdac:Work ;
	rdaw:P10002 <http://www.wikidata.org/entity/Q174596> . # has identifier for work > wd:Moby-Dick

Thanks to @pan-zhuo for pointing out this is not a correct use for this property. Why not? ... Well... Maybe it's easier to explain the relationship I intended to create during our discussion, and why (as I understand it) that isn't what the property is for.

In the implementation of 'has identifier for work' above, I was attempting to link descriptions of the same resource using the property. This is along the lines of what we've done in the UWLSWD Agents dataset, where we create a link between our description set for a given agent and the description set for the same agent in Wikidata by using owl:sameAs to assert that these entities are the same thing:

<https://doi.org/10.6069/uwlib.55.A.3.6#CurtisAsahel18741941> owl:sameAs <http://www.wikidata.org/entity/Q4803332> .

I understand this kind of relationship to be foundational for linking open data. We don't have to say everything about a thing. We can simply say what we know about a thing, which we can then unambiguously identify and, where known, link to external descriptions of the same thing. ('The power of linked data.')

But the property 'has identifier for work' -- and for that matter has identifier for expression, has identifier for manifestation, has identifier for item, has identifier for person, ..., ... -- is not intended to serve this purpose.

This property is intended to create a relationship between a Work (or an Expression, a Manifestation, etc.) and a Nomen. Not a relationship between the Work resource and an equivalent resource.

I'm still somewhat at a loss to explain exactly the difference between 'has identifier for [...]' elements and RDF properties such as owl:sameAs which create relationships between equivalent resources, but this is my best understanding at the moment, and I think this is made clear (as Zhuo helpfully pointed out) in the Toolkit guidance for using the IRI recording method with this property, (and also with the other properties linked above), which states:

Record an IRI for an instance of a Nomen as a real-world object.
Do not record an IRI for the entity that is being described.

* Perhaps important to mention, I think the guidance to record an IRI for a Nomen is meaningful in interpreting the property's semantics even in the case of the canonical--not datatype--property.

As a side note, I think it's interesting to consider the property label 'has identifier for [...]' with the usage of heading 'Identifiers' in the Wikidata web display, under which (it seems to me) properties creating relationships to equivalent resources are given.