After working through all of the possible options, we’ve decided to create our 
own class “Name” that we would propose be used in implementations for 
Linguistic Appellations (e.g. pretty much all of them that are not Identifiers, 
as far as we can tell).

The options, and the rationale for the decision:

* Multiple Instantiation:  We rejected this approach not because of the 
philosophy, but because a scan of implementations showed none that support it 
in practice. Yes, the assertion can be captured in a triple store, but once you 
try to do anything outside of merely recording it, implementations picked one 
of the classes (typically at random) for the programming level object to be an 
instance of.  This, to us, is a very strong indication that MI is an 
anti-pattern that should be avoided, and replaced with Multiple Inheritance to 
a single class.

* Associate multiple languages as literals (of datatype rdf:langString) with 
the Appellation. This would let us record the language along with the value. We 
rejected the approach as not allowing for valuable use cases, such as recording 
which name forms are translations of which, and being inconsistent with the 
patterns for Linguistic Objects. Consistency is very important when it comes to 
ease of use, and two patterns for pretty much exactly the same scenario is a 
costly exception. An additional consideration is that JSON-LD 1.0 doesn’t have 
great support for situations where there are both values that have languages 
and values that do not, generating unintuitive JSON constructions.

* Use a new predicate from Appellation for language, such as dc:language or 
schema.org:inLanguage.  Again, this would be entirely inconsistent with the use 
of P72 has language and create confusion as to when to use which pattern.  An 
additional consideration is that JSON-LD 1.0 doesn’t allow for the same key to 
be used to map to different predicates in different classes, which would let us 
at least hide the inconsistency from developers.  Both of these are fixed in 
JSON-LD 1.1, but it’s still early days in the W3C process for that and 
implementations don’t support them yet.

* Ignore the scope notes and use Title for all Linguistic Appellations. This 
would be both unintuitive and horribly inconsistent with systems that use Title 
following its official semantics. 

* Not associate language with Appellations at all. This is also a terrible 
non-solution for any international system.

* Introduce a new class, which is a sub class of both Appellation and 
Linguistic Object… which is what we’ve done.  The challenge is that systems 
that do not process the ontology will not realize that Name is really just an 
Appellation with the additional necessary functionality for languages. If there 
was a CRM ontology class, we would not need to do this.

For consistency, we will also create another class for the combination of End 
Of Existence plus Activity, to prevent the multiple instantiation problem there 
as well.  


Thanks,

Rob


On 9/12/17, 9:01 AM, "Crm-sig on behalf of Robert Sanderson" 
<[email protected] on behalf of [email protected]> wrote:

Yes, I understand that a single instance can have multiple classes – for 
example Destruction + Activity.  If that’s the recommended approach, that’s 
what we’ll work with.

    Thanks Martin!

    Rob

    On 9/12/17, 3:06 AM, "Crm-sig on behalf of martin" 
<[email protected] on behalf of [email protected]> wrote:

        Dear Robert,

        This is a basic question of modelling methodology, which must be 
clearly understood by those using any current formal ontology. We are modelling 
under the "Open World Assumption". This means that anything
        not said may be true or false, but never false per rule. If something 
is an instance of a class A, it does never imply that it cannot be an instance 
of a class B
        at the same time, except if we declare these classes as "disjoint".

        Hence, Appellation being not a subclass of Linguistic Object, does not 
mean that it is not a Linguistic Object. It means that an instance of 
Appellation
        may or may not be a Linguistic Object, and many other things not said. 
A Title on the other side is
        defined to be always a Linguistic Object.
        If an instance of Appellation, such as "information science" happens to 
be a Linguistic Object, you declare it to be instance of both classes, which is 
standard RDF/OWL. This is called "multiple instantiation"

        Title in the CRM has a more specific sense than just being a Linguistic 
Object and Appellation. We do not declare subclasses of combinations of classes 
just for the sake of an accidental combination.  It would fill the CRM with 
some thousand classes without
         particular meaning. You can do that for your own convenience in a 
local extension. If title were indeed regarded only an accidental combination 
of Appellation and Linguistic Object, it has to be deleted from the CRM.

        So, there is neither an intellectual nor technical issue to it, as far 
as I understand:-).

        All the best,

        Martin

        On 9/12/2017 12:50 AM, Robert Sanderson wrote:


        Dear all,

        Is there a reason why only Titles are also Linguistic Objects?

        The examples of Appellation:
           "the Forth Bridge"
         "the Merchant of Venice" (E35)
         "Spigelia marilandica (L.) L." [not the species, just the name]
         "information science" [not the science itself, but the name through 
which we refer to it in an English-speaking context]
         “安” [Chinese “an”, meaning “peace”]

        All of these Appellations are in a language (English, Latin, and 
Chinese) but only “The Merchant of Venice” can have its language explicitly 
declared, as Titles are also Linguistic Objects.

        If Appellation was a subclass of Linguistic Object (as a descendent of 
Symbolic Object) then this issue would go away. And also make Title just a P2 
of Appellation, rather than a special case.

        Alternatively, Title could be re-described to cover all Appellations 
with language, whereas E41 would be for Appellations like identifiers that do 
not have linguistic content.

        Rob




        _______________________________________________
        Crm-sig mailing list
        [email protected]http://lists.ics.forth.gr/mailman/listinfo/crm-sig



        --

        --------------------------------------------------------------
         Dr. Martin Doerr              |  Vox:+30(2810)391625        |
         Research Director             |  Fax:+30(2810)391638        |
                                       |  Email: [email protected] |
                                                                     |
                       Center for Cultural Informatics               |
                       Information Systems Laboratory                |
                        Institute of Computer Science                |
           Foundation for Research and Technology - Hellas (FORTH)   |
                                                                     |
                       N.Plastira 100, Vassilika Vouton,             |
                        GR70013 Heraklion,Crete,Greece               |
                                                                     |
                     Web-site: http://www.ics.forth.gr/isl           |
        --------------------------------------------------------------




    _______________________________________________
    Crm-sig mailing list
    [email protected]
    http://lists.ics.forth.gr/mailman/listinfo/crm-sig



Reply via email to