Hash: SHA1

Am 18.11.2014 um 20:35 schrieb Joseph Kiegel:
> Can someone supply bibliographic examples?
> How would situations like this be handled?

Are those real?

> bf:publication [ a bf:Provider ;
>            bf:providerDate "[1964]" ;
>            bf:providerName [ a bf:Organization ;
>                    bf:label "Bollingen Foundation" ] ;
>            bf:providerPlace [ a bf:Place ;
>                    bf:label "New York, N.Y. " ] ],

This is a container for a "publication event" pertaining to a particular
bf:publication. If we would assign an identifier it would not very much
increase usefulness because different resources would not have identical
publication events, only in the approximation that Bollingen Foundation
did it in New York about in 1964 - but for me this would constitute a
class of publication events, not an instance.

However the blank nodes for providerName (how can a *name* be an
Organization?) and providerPlace do have a very bad smell - even
in the bibliographic universe Bollingen Foundation and New York N.Y.
are recognized as entities

>        [ a bf:Provider ;
>            bf:copyrightDate "©1964" ] ;

this is an unrelated other event (as in actively registering the
copyright) also pertaining to the individual resource. No statement
is made about who the copyright claimant might have been. I would
interpret this not as a compicated way to code "1964" but as an
event where we cannot or do not wish to provide agents and places
connected with it. Thus again this registration event is completely
different from any other event labeled "copyright date 1964" for
different resources, and when crafting URIs for the blank nodes
they all *must* be different from each other.

> or
> bf:lccn [ a bf:Identifier ;
>            bf:identifierScheme "lccn" ;
>            bf:identifierValue "63010708" ] ;

We had a lengthy discussion about that some months ago, subsumed under
"real world identifiers": The URI
would be an (one) identifier for the bf:publication itself and RDF
does not give us the means to recognize it as an (the!) LCCN identifier.
Here we are asserting a much stronger statement: The bf:publication has
an LCCN identifier and its value is the string "63010708". (Some of us
happen to know that) this string is syntactically normalized and unique
to the publication and therefore can be used to identify resources.
Again we could craft a URI for the blank node, either a local one
faciliating comparisons within our dataset or - as also has been
proposed - even try to settle on a community-wide URI scheme which
would identify the identifiers and might open the door to specialized
web services acting on LCCN identifiers.

Another example (abridged from the german GND):

    a gndo:DifferentiatedPerson ;
    gndo:gndIdentifier "133655873" ;
    gndo:preferredNameForThePerson "Marchionatti, Roberto" ;
    gndo:preferredNameEntityForThePerson [
        gndo:forename "Roberto" ;
        gndo:surname "Marchionatti"
    ] ;
    gndo:dateOfBirth "1950"^^<> ;
    gndo:placeOfBirth <> ;
    owl:sameAs <> .

The gndo:preferredNameEntityForThePerson has a blank node as object,
acting as a container for first and last name of the person. I ~think~
that the FRANAR working group discussed at some point of time to make
each form of a name (as in see references) in authority records an
individual entity (and our "usual" authority entities then would refer
to a bunch of name entities spanning up the set of variant names and
reference forms) but since evolution did not take that path I'm quite
content with a blank node construction as stated above.

Unfortunately an only slightly more complicated example shows the
deficiencies of that approach:

    a gndo:DifferentiatedPerson ;
    gndo:gndIdentifier "131486888" ;
    gndo:preferredNameForThePerson "Marx, Carl Friedrich" ;
    gndo:preferredNameEntityForThePerson [
        gndo:forename "Carl Friedrich" ;
        gndo:surname "Marx"
    ] ;
    gndo:variantNameForThePerson "Marx, Carolus Fridericus", "Marx, Karl
Friedrich" ;
    gndo:variantNameEntityForThePerson [
        gndo:forename "Carolus Fridericus" ;
        gndo:surname "Marx"
    ], [
        gndo:forename "Karl Friedrich" ;
        gndo:surname "Marx"
    ] ;
    owl:sameAs <> .

There is a strong relation between "Marx, Carolus Fridericus" and
[ gndo:forename "Carolus Fridericus" ; gndo:surname "Marx" ] on one
hand and "Marx, Karl Friedrich" and [ gndo:forename "Karl Friedrich" ;
gndo:surname "Marx" ] on the other. Assigning identifiers to the blank
does not help us at all to make this relations explicit, for this we
would have to resort to completely different constructions like

    gndo:variantNameForThePerson [
        a gndxxx:nameInformationSet ;
        gndxxx:normalizedName "Marx, Carolus Fridericus" ;
        gndxxx:realLifePresentationOfName "Carolus Fridericus Marx" ;
        gndo:forename "Carolus Fridericus" ;
        gndo:surname "Marx"
    ], [
        a gndxxx:nameInformationSet ;
        gndxxx:normalizedName "Marx, Karl Friedrich" ;
        gndxxx:realLifePresentationOfName "Karl Friedrich Marx" ;
        gndo:forename "Karl Friedrich" ;
        gndo:surname "Marx"
    ] ;

(or even more complicated ones if we would start supplying sources for the
individual forms like LCNAF does or extend this model to corporate bodies
where acronyms and translated forms can be clustered alike). Here again I
would consider blank nodes as sufficient, at least as long we do not plan to
manipulate individual "variantNameForThePerson" subgraphs within the authority
record. But one might argue that even if /we/ don't intend to, some other
interested party very well might take advantage on individually adressable
variant forms within our authority data...

viele Gruesse
Thomas Berger

Version: GnuPG v1
Comment: Using GnuPG with Thunderbird -