I’m very supportive of the authorities proposal’s attempt to decouple authorities from the entities they describe. I hope that we can successfully maintain the semantic consistency in the data. I noticed throughout this thread (and in the VIAF data) authority URIs being used as the objects of properties where instead person, places, etc. URIs should be. For example, (as previously mentioned) VIAF has separate URIs for the person and the authority, but then they go on to make a sameAs assertion between the VIAF *person* and an id.loc.gov *authority*. I hope this is just a mistake, and that they would be open to making the sameAs assertion between the VIAF foaf:Document and the id.loc.gov authority.

For example: https://viaf.org/viaf/36997809/rdf.xml has a URI for the document ("http://viaf.org/viaf/36997809/“) and another for Carl Sagan himself ("http://viaf.org/viaf/36997809”, without the trailing slash). The Person URI has sameAs assertions with both authorities and entities modeled as persons.



Re:Roles, I’m a little confused about semantics of the role proposal, which may be related to Karen’s observation about bf:Instance data and bf:Agent data. Using the example in the proposal:

bf:contributor  [a bf:Contributor ;
	bf:role <some role resource> ;
	bf:agent <some agent resource> ].


I read this to mean some work (not represented in the example) has some bnode bf:Contributor. That bf:Contributor has a role. That bf:Contributor has a related agent. Is the bf:Contributor really more of a Contribution, like a PROV Activity (http://www.w3.org/TR/prov-o/)? If so, the proposal makes semantic sense, and I would just recommend naming bf:Contributor to bf:Contribution or just using PROV activity.

If the bf:Contributor is a bf:Agent then there are some issues. The bnode bf:Contributor (again, if it is a bf:Agent) is the *same as* the <some agent resource>; the <some agent resource> isn’t the *agent of* the bnode bf:Contributor. Also, if the bnode were ever replaced by a URI for the Agent, all other roles the Agent ever performed in other works would automatically be now associated with this work.

If a bf:Contributor is a bf:Agent it would make more sense just to say:

<some work> <some role/relator> <some agent resource> . 

(Note, the <some role/relator> property would have to have the range of an Agent. We would have to careful to not use properties that have domains of a work with a range of authority with bf:Contributors if they are agents.) 

The Agent in turn could be described by an authority:

<some work> <some role/relator> <some agent resource> .

<some agent resource> <bf:identifiedBy> <some authority> .



Re: the Identifiers proposal, Karen has said this before, but we shouldn’t conflate the role of URIs with string identifiers. When I read the Authorities proposal, I interpret bf:identifiedBy as having the semantics:

<some thing> <is identified through an authority> <some authority resource> .

**But** the identifier proposal says the bf:identifiedBy property is for an identifier string. That’s very different. I would recommend we be able to say the following through different properties:

<some bf:Resource> <is described by an authority> <some authority resource> .
<some authority resource> <is identified by an identifier> <some string identifier> .


Thank you for your consideration,
Steven

————
Steven Folsom
Metadata Strategist and Standards Advocate
Cornell University Library







On 8/25/15, 5:58 PM, "Bibliographic Framework Transition Initiative Forum on behalf of Thomas Berger" <[log in to unmask] on behalf of [log in to unmask]> wrote:

>-----BEGIN PGP SIGNED MESSAGE-----
>Hash: SHA1
>
>Am 25.08.2015 um 19:58 schrieb Karen Coyle:
>
>>> bf:contributor [
>>>    a bf:Contributor ;
>>>    bf:role<http://id.loc.gov/vocabulary/relators/ill>  ;
>>>    bf:role<http://id.loc.gov/vocabulary/relators/aut>  ;
>>>    bf:agent<http://id.loc.gov/authorities/names/1234>  ;
>>>    bf:agent<http://d-nb.info/gnd/9876>  ].
>>>
>>> (MARC semantics: two different roles for one bibliographic
>>> identity accounted for in different authority files)
>> 
>> Wouldn't it be preferable that any equivalence between agent identifie
>rs
>> be declared outside of the instance data for a bibliographic resource?
>> Logically, the information that they are equivalent is not related to 
>a
>> particular bibliographic resource, but is a fact related to the agents
>> themselves. It should hold true for all instances.
>
>Well, I wouldn't expect that one cataloging agency would tie its
>activities to more than one authority control environment. And if,
>it probably would encapsulate that into some local construction
>detailing on the exact findings of the agency:
>
>bf:contributor [
>   a bf:Contributor ;
>   bf:agent<http://example.com/authorities/xyz> ]
>
><http://example.com/authorities/xyz> [
>  a schema:Person ;
>  a bf:Agent ;
>  rdfs:label "John Doe (1900-1983)" ;
>  skos:exactMatch <http://id.loc.gov/authorities/names/1234> ;
>  schema:sameAs <http://d-nb.info/gnd/9876> ;
>  owl:differentFrom <http://www.wikidata.org/entity/abc>
> ].
>
>But for some bibliographic aggregator who does not want to presuppose
>the authority control ecosystem potential consumers of its data would
>prefer I can imagine descriptions where alternative(!) identifications
>for the bf:contributor are supplied. I mean, there certainly must have
>been a use case when MARC's 7XX $0 was introduced as repeatable?
>
>I'd consider this "instancial" evidence of possible identity, declaring
>equivalence in the context of a bf:Agent would be a stronger statement.
>
>BTW: Different authority control systems use rather different
>vocabularies to establish cross border identification of their
>resources (there seems to be a general tendency to avoid owl:sameAs,
>though). Given the fact that even for simple agents like persons
>the entities not always are in 1:1 correspondance (e.g. LCAuth focussing
>on bibliographic identities, and ULAN very much on real persons),
>could or should there be a bf:identicalTo property declaring equivalence
>in a Bibframe sense? Or would identity rather be something to be
>defined by the application, e.g. data created in the context of a
>manuscript and letters application would probably not distinguish
>between Ruth Rendell and Barbara Vine and thus would use "regular"
>authority control with a deviant notion of equivalence ;)
>
>
>> (This brings up the question of usage of VIAF identifiers in the place
>> of specific authority control identifiers, but that's another topic.)
>
>Maybe not at all. Certainly VIAF is lacking some features we are
>accustomed to find in more traditional authority control systems,
>but if some people successfully start to use VIAF in lieu of
>ordinary authority control, then my expectation on Bibframe would be
>that VIAF based authority control should not work differently from
>ordinary ones. So "VIAF" could serve as kind of litmus test for
>Bibframe concepts ans properties - if they rely on "traditional"
>features not present in VIAF they are lacking universality (in the
>universe Bibframe is concerned with) and should be considered optional
>at most.
>
>viele Gruesse
>Thomas Berger
>-----BEGIN PGP SIGNATURE-----
>Version: GnuPG v1
>Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/
>
>iJwEAQECAAYFAlXc5QMACgkQYhMlmJ6W47NXIQP+OQoHIL6caieYMysDVvGrilf5
>jHAo//n7tbH2T16iR5aoy8C4WO8rtJFgFvLPfJH22P5b2QhAA3oeHctqAMi/P8gk
>5X48gAwqYrQffDw6glSD+XL6a59p0JIUQFWW4lFpEq8Y+dLp3Jz9Kq4VCkDMDsul
>Jht1oMxqC+v+azrB9Hs=
>=Kcxq
>-----END PGP SIGNATURE-----