On 5/16/13 5:54 AM, Owen Stephens wrote:
[log in to unmask]" type="cite">
On 16 May 2013, at 12:16, "Meehan, Thomas" <[log in to unmask]> wrote:

Given that a) BIBFRAME cannot possibly control all relationship designations and b) BIBFRAME needs to handle both URIs and textual relationships, I cannot see how this can possibly be kept simple- although I dearly hope it can be- without being willing to lose some data. The current BIBFRAME approach of putting the relationship in the Authority can't work but I wonder, reluctantly, whether a BIBFRAME Relationship entity (or a fancy blank node) standing between the Work and Authority is the only way to do it 
I think you are right. Also worth looking at the recently published "Scholarly Contributions and Roles Ontology" ( SCoRO) at http://semanticpublishing.wordpress.com/2013/04/19/scoro/ and also the related "Publishing Roles Ontoloty" (PRO) at http://purl.org/spar/pro/ 

These ontologies share some of the concerns of BIBFRAME but in a specialist arena (scholarly articles) - but they show that having an additional layer to express roles in the creative process is not without precedent.

Owen, the way I read SCoRO, there is an entity for the person that is the subject (left hand side of the triple) of all of the roles (in the place of defining the role in the bibliographic entity). Within the person entity, roles then point to the works. This is actually something that some of us have proposed as an authority format in the past -- that the author authority entry should have links to the works.

SCoRO has some very complex relationships to handle and their structure is therefore pretty deep. However, in fact what I believe they have essentially done is switch the triple:

BookA -> has author -> PersonB


PersonB -> isAuthorOf -> BookA

This means that the link between the person and the book has the person as the subject, rather than the book:

PersonB   foaf:name  "John Smith"
PersonB   bf:editor  WorkABC
PersonB   bf:author  WorkXYZ
PersonB   bf:author  Work123

This is a more person-centric view than library data has exhibited, and it makes sense that this view would come out of the academic community (which is SCoRO's emphasis), where who did what in scientific work is key. This does seem to make it easier to ask questions like: "what are all of the works translated by...?" although one should be able to obtain the same result using library data where the role is coded with the bibliographic item as the subject, rather than the person as the subject. (Note, also, that programs could create the added triples so that" A is author of  B" is also represented as "B has author A," giving you the best of both worlds.)

It still doesn't solve our problem of roles without URIs, which cannot be in that middle position in the triple.
[log in to unmask]" type="cite">

Because :creator, :composer and the agent involved are all objects of the triples, they could easily be literals or URIs (and even those URIs can't possibly always be BIBFRAME ones).
I don't understand the last part of this statement - what would it mean for the URIs to be 'BIBFRAME' ones in this case? I think we could expect a set of data to be self-sufficient (i.e. to coin URIs where required and not rely on external entities if the desire is to avoid this)

I expect the last thing BIBFRAME wants to do is add another first class entity to the model and querying the above would be considerably less fun, but given the willingness to apply complexity to Annotations and the importance of relationships but their wide variance in past and anticipated practice (again, RDA app. I), is this a silly idea? I did wonder at one point whether this additional layer was what BIBFRAME Authorities was aiming at anyway.

I think this is a good idea. If we want to express creator/contributor relationships in a sophisticated way we have to expect some complexity, but I don't think this approach is really that bad in terms of complexity - and of course it is worth repeating that if one was actually cataloguing an item, the user interface should be simple, or at least easy to use, not matter how sophisticated the underlying data model.

I wonder if there isn't some confusion between the BIBFRAME graph and the sources of the data in that graph. BIBFRAME can declare an authority entity with, say, a display form of the name and a bit of biographic data, but that doesn't that the data itself "belongs to" a BIBFRAME specific declaration, nor that it is created redundantly for BIBFRAME. The data can be any RDF resource. This means that a BIBFRAME authority entity can contain a display label from LCNA (or DNB, etc.), and biographical data from DBPedia. It would look like:

   bf:label -> LCNAn3939393 -> {?query -> madsLabel ->} {result: "Barbara Cartland"}
   bf:bio ->   DBpediaBC -> {?query dbp:bio ->} {result: "Born in ....."}

Since graphs are formed through queries, you can obviously cache the result of your query locally to avoid re-querying each time, although you then need a method to update your query on an appropriate schedule.

In this way, the BIBFRAME graph is part of the growing graph of data, not a separate entity.


[log in to unmask]" type="cite">

I know what you mean about wondering if this additional layer was what BIBFRAME Authorities were aiming at - but my conclusion is that they are there for a different purpose ... working on an email on this now :)


Karen Coyle
[log in to unmask] http://kcoyle.net
ph: 1-510-540-7596
m: 1-510-435-8234
skype: kcoylenet