Print

Print


Great points, and thanks for the discussion as always, Karen. 

For some background, we have been using an RDF-based schema/ontology for 
a Hydra-based digital repository for almost 5 years. On the other 
thread, there was a lot of guessing about what linked data can do, and I 
think I'm in a fortunate position to say what a non-schema.org, RDF-
based library schema can pull off. 

Even without schema.org properties we achieve good SEO (e.g. someone can 
type in a reasonably proximate collection name in Google and we are the 
top 3 results). But as you cogently pointed out elsewhere, simple things 
like adding maps become incredibly complicated, because all the 
wonderful tools built around the major APIs tend to assume JSON data and 
an entire stack of technologies different than ours. Geospatial metadata 
simply has no RDF options (we are hoping along with DPLA for 
developments in geoJSON-ld and other schemes), and we are stuck waiting 
for that. The solution in the meantime has been to kludge it so that we 
have basic Leaflet integration. So a relevant question is, who will 
build this whole new marketplace of RDF-based apps that aren't just 
parsing or serialization libraries? If we are only using this data that 
has to be kludged just to support existing apps/APIs, there is little 
value added. 

What our metadata unit has most crucially learned is that making your 
own predicates (or properties) is potentially dangerous because now you 
have lost the power of the crowd to improve understanding or 
implementation of it, or even if the predicate is necessary in the first 
place. Someone on another thread said RDF-based data is instantly able 
to be integrated with one another. In theory, yes, but in practice, no. 
It's true that blind trust of others' schemes is necessary and you could 
merge any RDF graph with another, but we still have to evaluate where a 
data source fits into ours, what their domain/ranges are for predicates, 
etc.

When I look at BIBFRAME, I see an entire ontology built anew. Why wasn't 
a first principle to borrow or at least map to as many Dublin Core, 
PREMIS, etc. predicates as possible? Is that a goal and a known task, 
and I shouldn't worry about it? If it doesn't at least map to existing 
predicates used in the linked data world, it will become a niche 
ontology. 

Our new ontology will have very little BIBFRAME predicates because it 
doesn't seem to make any sense with its insistence on the 
Work/Instance/Item framework. Digital objects are just that... objects, 
and we don't think our objects fit into this paradigm. That said, we do 
sometimes have objects that come from our catalog, and if BIBFRAME ever 
comes to our library we would of course figure out a mapping to our 
ontology, that would be much easier than the current MARC->MODS->our 
ontology mapping. But that's rather disappointing in the end. 



--Ryan



On Thu, 2 Feb 2017 13:21:56 -0800, Karen Coyle <[log in to unmask]> wrote:

>Simon, interestingly, this person[1] thinks that BIBFRAME is too much 
of
>a conceptual model.
>
>One of the problems that I have is that the actual vocabulary doesn't
>always match the stated concepts. I had this experience recently with
>the Zepheira version, bibfra.me [2], that people often call "BF-lite".
>For example, "Agent" class includes the property "audience" from
>Resource, but the definition of "audience" talks about the content of a
>resource. Things like that just jar me and don't seem logical. We went
>through that with the earliest version of bibframe that didn't seem to
>put things at the correct bibliographic level. A lot got fixed, it
>seems. However, it's hard to judge the model when the execution brings
>up questions, and when the model is defined in just a few sentences.
>(Note: schema.org has many of the same problems, but there's a large
>community that discusses them so one has hope that they'll eventually 
be
>worked out.)(Also note: FRBR has this same problem with its user tasks,
>that are covered in one paragraph each, and to me are totally vague.)
>
>Admittedly, FRBRoo is much more rigorous. That said, it needs something
>like the BF-lite presentation to make it understandable. The best thing
>about BF-lite is its web site organization and presentation.
>
>Also, shouldn't we be creating standards using methods like are used by
>W3C and IETF - with open communities, wikis, mailing list archives, 
open
>documents? That doesn't mean that organizations develop a standard and
>then post it online, it means that the PROCESS needs to be visible so
>that people can participate, or at least understand the end result. 
It's
>very hard to understand a standard if you haven't seen what was
>discussed, what was dismissed, what the thinking was. We're way behind
>others in our standards process.
>
>kc
>[1]
>https://redlibrarian.github.io/article/2017/02/01/library-systems-
disaster.html
>[2] http://bibfra.me/
>
>On 2/2/17 12:24 PM, Simon Spero wrote:
>> On Feb 2, 2017 7:21 AM, "Gordon Dunsire" <[log in to unmask]
>> <mailto:[log in to unmask]>> wrote:
>>
>>     "I do not understand why RDA cataloging examples and 
implementations
>>     have not picked up Bibframe as a prerequisite. They seem like not
>>     being made for each other, which is confusing and kind of 
bizarre.":
>>     I think the second point is answered earlier in the paragraph: 
"It
>>     is so simple that it even does not follow FRBR ..."
>>
>>     There are other reasons why RDA does not regard BIBFRAME as a
>>     prequisite:
>>
>>     It is not stable.____
>>
>>     Its functional requirements are unclear.
>>
>> Quite.
>>
>> To the extent that BIBFRAME's functional requirements /are/ clear, 
your
>> remarks above are not signs of success. Remember that the goal of the
>> bibframe effort was set by the LC report on the RDA test, and it's
>> purpose was to establish a non MARC based approach for carrying RDA
>> data. The report did not call for establishing a new conceptual 
model,
>> and this may have been unwise, and contributed to the instability 
noted.
>>
>> A different  starting point would have been to start from the basis 
of
>> FRBRoo, which /is/ a rigorously defined FRBR based model, and define 
any
>> simplified or extended ontology in alignment with that. Such an 
approach
>> would also consider and make explicit the functions that a less 
record
>> based approach could benefit, and what sort of enabling workflows,
>> infrastructure, and architecture might be needed to support those 
goals.
>>
>> Simon
>
>--
>Karen Coyle
>[log in to unmask] http://kcoyle.net
>m: +1-510-435-8234
>skype: kcoylenet/+1-510-984-3600
>=======================================================================
==