Comments inline.
Yours,
Kevin
On 07/28/2014 06:21 PM, Robert Sanderson wrote:
>
> Hi Kevin,
>
> On Mon, Jul 28, 2014 at 1:53 PM, Kevin Ford <[log in to unmask]
> <mailto:[log in to unmask]>> wrote:
>
> I believe Rob is trying to underscore the fact that there are
> variable ways to record a Work's title (not to mention an
> Instance's) and, because there are variable ways to do it, the query
> becomes, well, ridiculous.
>
>
> I didn't say ridiculous, but I did think it many times, and am glad that
> you did :)
>
>
> So, I believe Rob kicked off a thread that basically asked why are
> there two methods to capturing title information.[1] One way is to
> use a literal string and the other is to reference a bf:Title
> resource. [... examples ...]
>
> The variability probably reflects a few things: trying to have it
> both ways so that implementers have the option; a misunderstanding
> or miscommunication about how multiple groups and individuals think
> about "titles;" a desire to accommodate old and new cataloging
> rules; and a level of parity with current cataloging practice.
>
>
> The first is the dangerous one, and particularly so in a graph. In my
> opinion, the model should pick one. Whichever is the simplest case that
> covers the use cases. If that's string literals, fantastic. If not, I
> could live with blank nodes... just not both at once.
I hear you. I worry about two things: 1) the prevalence of anonymous,
largely not-reusable resources in general, and the impact that could
have at scale, and 2) designing for uncommon cases.
About 2: They're not "corner cases" but I wonder if the vast majority of
titles (and by "vast" I mean certainly greater than 90% of them)
associated with Works require the use of a relationship /and/ a resource
to capture the title. Said another way, would we end up with more than
90% of our Works looking like this
ex:1 a bf:Work
bf:title _:x
_;x a bf:Title
bf:value "The Scarlett Letter"
versus
ex:1 a bf:Work
bf:title "The Scarlett Letter"
?
I offer the above simply for comment - don't misunderstand, I get the
problem with there being two ways to do something. I'm just wondering
if there is a way to have our cake and eat it too.
>
> Personally, I am in favor of investigating treating titles as string
> literals. (And I mean all titles: constructed titles, regular
> old-fashioned-this-is-the-__title titles, abbreviated titles, spine
> titles, key titles, added title page titles, etc.)
>
>
> _:Work bf:(constructed|abbreviated|spine|key|added|etc)Title "literal" .
> or:
>
> _:Work bf:title [ a
> bf:(Constructed|Abbreviated|Spine|Key|Added|Etc)Title ; bf:value
> "literal" ] .
>
> I prefer to multiply classes rather than predicates but the first way is
> much more straight forward if the only property of the Title resource is
> its value.
>
> However, that's a lot easier to say than it is to robustly test and
> robust testing is needed because titles clearly unearth a number of
> little-considered but real issues, such as a Work with multiple titles,
>
>
> _:Work bf:title "literal 1" ;
> bf:title "literal 2" .
>
> each with translations or transliterations [2]
>
>
> This becomes trickier, obviously.
>
> It seems like you would need to have a resource to express
> translation/transliteration relationships between.
>
> _:Work bf:title _:title1 .
>
> _:title1 a bf:AddedTitle ;
> bf:value "literal" ;
> bf:language lingvo:en ;
> bf:hasTranslation _:title2 .
>
> _:title2 a bf:TranslatedTitle ;
> bf:value "littéral" ;
> bf:language lingvo:fr .
>
>
> and the need to capture the fact that a title was added by a cataloger
>
>
> If this sort of provenance is needed at the title entry level, either
> the assignment is reified (see also Simeon's examples in the Identifier
> thread, and also bf:Annotation) or there would have to be a resource
> with an assigner predicate.
Roy Tennant's marcusage app is not working presently, but I wanted to
know exactly how often something like MARC BIB 242 (translation of title
by cataloging agency) is actually used for titles. I've started to
wonder if reification in these types of circumstances might not be the
better way to sequester the complicated things, to be used only when
needed (which is to say, rarely, at least in the grand scheme of things).
I mean titleAssigner to follow current
> conventions ;) [please please just assigner!]
>
> _:title2 a bf:TranslatedTitle ;
> bf:value "littéral" ;
> bf:language lingvo:fr ;
> bf:translatedBy <http://translate.google.com/> .
>
> And bf:translatedBy should probably be rel:trl
>
> While the bf:Title construct exists as an attempt to address /some/
> of those less common cases (such as a cataloger assigned titles), it
> remains problematic because it is hard to square that particular use
> case with existence of "bf:formDesignation" or "bf:titleAttribute,"
> the definitions of which strongly suggest they are aspects of the
> Work, not a "title."
>
>
> +1. These should be taken off of title and pushed up to the Work.
>
> Which brings me to another point: When we - as listserv
> participants - say "Title" are we always talking about the same thing?
>
>
> To be clear, I mean the name of the Work. Or its label. Its dc:title.
>
> So, I think these last questions are the first ones we need to find
> agreement on.
>
> 1) Is a title an attribute or property of a Work or Instance? Do
> you think of a "title" as synonymous with a Work (or Instance), that
> is, the thing you are describing?
>
>
> A property (or relationship if the title needs to be Thingified)
>
> OR
> 2) Is a title a type of Thing unto itself, one that can have its own
> identifier, and is related to but otherwise distinct from the Work
> or Instance you are describing? It is something that is associated
> with a Work but is not necessarily a property or attribute of the
> Work? Though this is not only way to look at this, one wants to
> ask: Are titles re-usable?
>
>
> It might be a thing unto itself, if it is important to capture
> information regarding its provenance or other features.
>
> For example, if it was important to capture that Rob Sanderson
> translated the title, then there was a TranslationActivity that occured
> at a certain point in time performed by Rob with the English title as
> input and the French title as Output.
>
> Overkill? I think so... but that's the implication of the translated
> title use case where there's more information to be associated with the
> title separately from the Work/Instance.
I agree about it being overkill, and I'm now wondering how realistic,
and common, some of these use cases are. I'm not against uncommon use
cases, but in re-reading all of this just now I started feeling like the
agenda was being set by uncommon use cases. In reviewing the above, and
looking at the BF Vocabulary, I'm not clear about the intent of
titleSource.
"titleSource" - I think - has an association with MARC 210 $2, which
references this MARC source code list:
http://www.loc.gov/standards/sourcelist/abbreviated-title.html
Suddenly, those seem like candidates to become properties to be
published at ID.LOC.GOV, perhaps, and employed in the place of
titleSource.
There's a number of simplifications I think and I'm looking to tease
those out.
Yours,
Kevin
>
> And yes, titles should be reusable if they're resources.
>
>
> Rob
>
> --
> Rob Sanderson
> Technology Collaboration Facilitator
> Digital Library Systems and Services
> Stanford, CA 94305
|