On 5/16/13 2:15 PM, Ford, Kevin wrote:
>> To resolve this, and this is a case that we will encounter, there will
>> have to be some contortions, possibly a blank node:
>>
>> WorkURI - Creator -_blankA
>>
>> _blankA - typeOf - http://bibframe.org/vocab/Person _blankA - authID -
>> http://bibframe.org/auth/person/franklin
>> _blankA - http://bibframe.org/vocab/label - "Franklin, Benjamin, 1706-
>> 1790"
>> _blankA - http://bibframe.org/vocab/resourceRole - "supposed compeser"
> -- This solution is, for all intents and purposes, identical to the discussion paper's Example 3 [1].
Kevin, if it is identical, then it points out that this graph is NOT
re-usable. Or at least, re-use is extremely unlikely. (I used a blank
node specifically because it generally represents a non-reusable graph,
a one-off.) What this does is it creates a graph that represents a
person in a specific relationship to a bibliographic entity. It is a
"one-to-one" graph. That was what I intended to model.
>> This brings up a specific question about the BIBFRAME authority: Is it
>> intended to be re-usable? Or does it have the "one-off" nature of a
>> blank node?
> -- It is designed to be re-usable. For example, I would expect an implementer to create only one BIBFRAME Authority per Agent entity. Now, what this email and a few others I've read have made me realize is that implementers will have to create duplicate BIBFRAME Authority resources for what is essentially the same entity in some special cases. For example, if it is necessary to capture the "resourceRole" as free text,
Kevin, that is such a bad idea that I think we need to look seriously at
alternatives, which do exist. The SCoRO design is one possible
alternative, albeit not "lightweight." Alternatively, a blank node added
to the bibliographic graph could model this data. But by all means, it
is not going to be a good idea to create multiple graphs for the same
Agent, and especially if this graph represents an authoritative identity
for the entity. Think about all of the links that could relate to that
person: alternate names (which still haven't been addressed), bios,
images, DBPedia, various IDs (VIAF, ORCID), etc. You would have to link
those to each graph. I don't think that's tenable.
> such as in the Bad Data example, the implementer will have to basically create two BIBFRAME Authority resources (BFA1 and BFA2) for Benjamin Franklin: BFA1 would not contain a "resourceRole" property and it could be used whenever the relationship between the Work and Benjamin Franklin's role was known (creator, illustrator, etc), but BFA2 would have the resourceRole property and would be associated with one distinct Work. It would be my hope that someone eventually recognize the problem and do away with BFA2 by creating the triple "W :cmp BFA1." Unless, of course, the "supposed" is deemed an essential aspect, at which point BFA2 would have to continue to exist. Something else to keep in mind: this example may not quite meet the definition of "edge case" but it also unlikely represents a significant percentage of cases to demand a more elaborate across-the-board solution.
First, I don't think this is an extremely rare edge case. It *is* one of
the disadvantages of seeing BIBFRAME as the recipient of MARC data as
opposed to a data description in its own right. Yes, there will be edge
cases and bad data -- mainly because many MARC creation systems didn't
do much quality control. But I would endeavor (:-)) to tuck the
"MARC-ish-ness" someplace where it doesn't do any harm. My first
thought was that we should just create a property called "originalMARC"
where we would stuff the entire MARC record for future reference, much
like the current BIBFRAME stores the fixed fields, unchanged. That's
probably a pipe dream, and we'd have to decide whether to link that to
the Work or the Instance, or both. But there are surely options that
don't result in multiple URIs for what we know is the same Agent entity.
kc
>
> Personally, I think the "resourceRole" property should be avoided at all costs. It exists to ensure that data, such as visible in the "Bad Data" example (it's a real example, incidentally), not be lost in a migration. It also exists to accommodate RDA's allowance of a write-in option. I expect future cataloging practices (to be driven by new cataloging interfaces) will encourage the use of a property to define the relationship between an Agent and a Work.
>
> Yours,
> Kevin
>
>
> [1] http://bibframe.org/documentation/bibframe-authority/#issues-3
> [2] http://bibframe.org/vocab/resourceRole
>
>
>
>> -----Original Message-----
>> From: Bibliographic Framework Transition Initiative Forum
>> [mailto:[log in to unmask]] On Behalf Of Karen Coyle
>> Sent: Tuesday, May 14, 2013 8:36 PM
>> To: [log in to unmask]
>> Subject: [BIBFRAME] BIBFRAME authorities: relationToWork
>>
>> There is an example under "Bad Data":
>>
>> <!-- BIBFRAME Work -->
>> <*Play* id =
>> "*http://bibframe/work/Quatuor-pour-trois-violons-et-violoncelle*">
>> <title>Quatuor pour trois violons et violoncelle</title>
>> <creator resource = "http://bibframe/auth/person/franklin" />
>> </*Play*>
>> <!-- BIBFRAME Authority -->
>> <*Person* id="http://bibframe/auth/person/franklin">
>> <label>Franklin, Benjamin, 1706-1790</label>
>> <relationToWork>supposed compeser.</relationToWork>
>> <hasIDLink
>> resource="http://id.loc.gov/authorities/names/n79043402" />
>> <hasVIAFLink resource="http://viaf.org/viaf/56609913" />
>> <hasDNBLink resource="http://d-nb.info/gnd/118534912" />
>> </*Person*>
>>
>> There is a problem with "<relationToWork>supposed
>> compeser.</relationToWork>" that goes beyond the bad data.
>>
>> I believe that someone else mentioned this, but if "relationToWork" is
>> a property of the Person, then that Person ID cannot be used for any
>> other relationship (e.g. "author") because as expressed here
>> relationToWork is being said about the Person, period. There needs
>> instead to be a triple that has:
>>
>> WorkURI - relationToWork - PersonURI
>>
>> This then states the relationship of the Person to that particular work,
>> and doesn't change the underlying definition of the Person.
>> Unfortunately, this isn't possible when the value of "relationToWork"
>> is not a URI, because properties (the middle parts of the triples) have
>> to be URIs. For example, if this information were coded using the MARC
>> relator standard "<http://id.loc.gov/vocabulary/relators/cmp>" then you
>> could say:
>>
>> http://bibframe.org/work/Quatuor-pour-trois-violons-et-violoncelle -
>> http://id.loc.gov/vocabulary/relators/cmp -
>> http://bibframe.org/auth/person/franklin
>>
>> Which means that this person is the composer of the Work. (I don't know
>> if there's a code for "supposed composer" but I'm pretending that "cmp"
>> will do.)
>>
>> To resolve this, and this is a case that we will encounter, there will
>> have to be some contortions, possibly a blank node:
>>
>> WorkURI - Creator -_blankA
>>
>> _blankA - typeOf - http://bibframe.org/vocab/Person _blankA - authID -
>> http://bibframe.org/auth/person/franklin
>> _blankA - http://bibframe.org/vocab/label - "Franklin, Benjamin, 1706-
>> 1790"
>> _blankA - http://bibframe.org/vocab/resourceRole - "supposed compeser"
>>
>> The difference between the blank node and the way I interpret the
>> "lightweight BIBFRAME authority" is that the blank node is a one-off --
>> it's only usable in this one instance. There is no creation of an
>> identity that could be re-used in other circumstances (although, as
>> I've done above, it can link to an identity, such as an authority
>> record).
>> Blank nodes are considered last resorts by some linked data enthusiasts
>> because they aren't re-usable. But there are times when you really have
>> little choice.
>>
>> This brings up a specific question about the BIBFRAME authority: Is it
>> intended to be re-usable? Or does it have the "one-off" nature of a
>> blank node?
>>
>> kc
>> p.s. There are undoubtedly other solutions to this problem, and I hope
>> they'll get posted here.
>>
>> --
>> Karen Coyle
>> [log in to unmask] http://kcoyle.net
>> ph: 1-510-540-7596
>> m: 1-510-435-8234
>> skype: kcoylenet
--
Karen Coyle
[log in to unmask] http://kcoyle.net
ph: 1-510-540-7596
m: 1-510-435-8234
skype: kcoylenet
|