W3C home > Mailing lists > Public > public-lld@w3.org > October 2011

Re: Antw: Linked Library Holdings/Items

From: Adrian Pohl <pohl@hbz-nrw.de>
Date: Mon, 17 Oct 2011 14:39:23 +0200
Message-Id: <4E9C3E1B020000140004256F@agrippa.hbz-nrw.de>
To: <public-lld@w3.org>
Hello Jakob,

 >>>Jakob Voss <jakob.voss@gbv.de> schrieb am 17.10.2011 um 13:09:
> Hi Adrian,
>>> I just started to experiment with RDF encoding of information
>>> about library holdings. Has anyone done something in this
>>> yet, so we can adjust our conceptual models and ontologies?
>> We have done something in this direction with lobid.org. Currently,
>> we only link frbr:manifestations (the non-series things described
>> library records) to organisations which have at least one item
>> instantiating this manifestation. We don't re-use or mint any item
>> URIs and, thus, use blank nodes.
>  >
>  > http://lobid.org/resource/HT002948556
> Good to check. An item URI for the same book in one of our libraries
> http://uri.gbv.de/item/opac-de-18:epn:354541463
> so we only need to somehow map both via frbr:exemplar :-) 

Yes, the only thing we can do at the moment (as you don't have GBV-URIs
for the manifestation) is link the lobid entry to the GBV exemplar:

<http://lobid.org/resource/HT002948556> frbr:exemplar

> However I 
> currently focus on items, because I don't know what a manifestation 
> really is (especially in contrast to frbr:Expression) and because I 
> hesitate to create new URIs for thins that already have URIs. I could

> create an URI for the same Manifestation, for instance
> http://uri.gbv.de/document/gvk:ppn:16523315X - the book itself
> http://uri.gbv.de/record/gvk:ppn:16523315X - catalog record of the
> And say
> <http://lobid.org/resource/HT002948556> owl:sameAs
> <http://uri.gbv.de/item/opac-de-18:epn:354541463>

You mean: <http://lobid.org/resource/HT002948556> owl:sameAs
<http://uri.gbv.de/document/gvk:ppn:16523315X > .

> But how about URIs for the same Book in WorldCat and other catalogs?

> Should everone mint his own URIs for FRBR expressions and

Everyone has created his own records describing FRBR manifestations
until now... A decentral approach seems sensible to me especially
regarding performance and issues of long term preservation. We then have
to figure out how edits to a record are communicated between instances.
(In hbz-internal discussions Jörg Prante has proposed using a chat
protocol like XMPP for this.)

> By the way it would be nice to enrich your example with some links to

> external sources, for instance:
> http://www.worldcat.org/oclc/630937301
> http://www.librarything.com/work/143122

I think you are getting into problems if you do this automatically as
you can't be sure whether the thing described in WorldCat, LibraryThing
etc. is really the same. (You could of course use another predicate than
owl:sameAs but we have to think about use cases first to find out what
links really make sense.) Also, this is what actually is going on in the
culturegraph project and we hope to get good information where to link
to out of it in the coming months.

>>> In lack of existing RDF properties for library holdings I created
>>> http://purl.org/ontology/daia/label
>>> for call numbers and
>>> http://purl.org/ontology/daia/heldBy
>>> for a "holding" relationship between item and a library
>>> institution. (by the way: is "held by proper English or better
>>> "hold by"?). Unfortunately the international ISIL agency has not
>>> defined an URI prefix for ISIL yet, so I recommend to use the
>>> prefix from lobid.org.
>> The holding relationship can also be represented by
>> http://vocab.org/frbr/core#owner ("A property representing an
>> that owns an item."). So, you don't have to mint a new one...
> A library does not necessarily own a book that it holds. Same applies
> digital documents where a library has licensed access to it. So I
> frbr:owner to narrow in scope. daia:holds / daia:heldBy was created
> express the relationship between a book and a library institution
> holds this book it its collection, so I don't have to think about 
> difficult questions of legal right of property and ownership.

I understand. I just don't put this legal semantics into this
predicate... Generally, I found Francisca's mail interesting, refering
to CIDOC where an item obviously would be linked to a collection instead
of directly to an organization. That make ssense as big libraries may
have many different collections at different places. Relating an item
primarily to a collection would mean adding useful information about
location, access etc. But you can do this afterwards, anyway...

>>> The FaBiO has some more properties that could be used to describe
>>> library holdings, but I hesitate to reuse purely theoretical
>>> ontologies from academics that have not been applied to real world
>>> library data.
>> Same problems here. We don't use FaBiO because it isn't used
>> anywhere else. Also it is quite complex and implementing some kind
>> idiosyncratic FRBR flavor (who doesn't...). It doesn't seem very
>> sensible to me in some parts I've looked at.
> ok, maybe I'll drop it too.
>>> Do we have URIs for "ONIX media format" and "MARC media format"
>>> and how do you relate an item to its format?
>> Until now, we have used literals for specifying the format and
>> using the properties dcterms:format and dcterms:medium. It's all
>> very stable yet.  Input for the literal is MAB field 050[1]. We'd
>> also be happy about a controlled vocabulary.
> I found a good description by Eversberg (in German):
> http://www.allegro-c.de/regeln/rda/chap3.htm

Thanks, I will have a look at it.

> In RDA we have:
> 1. media type
> 2. carrier type
> 3. content type
> There are (unstable?) URIs at
> http://metadataregistry.org/concept/list/vocabulary_id/37.html
> http://metadataregistry.org/concept/list/vocabulary_id/46.html
> http://metadataregistry.org/concept/list/vocabulary_id/45.html
> @prefix rdamedia: <http://rdvocab.info/termList/RDAMediaType/>
> @prefix rdacarrier: <http://rdvocab.info/termList/RDACarrierType/>
> @prefix rdacontent: <http://rdvocab.info/termList/RDAContentType/>
> Is there consensus which RDF properties to use for relating FRBR 
> resources and RDA media/carrier/content types?

I don't think so but would like to hear more about this.

>>> I am still looking for good RDF properties to connect an item to
>>> the library record which is about the item's frbr:Manifestation
>>> (or frbr:Expression?).
>> Why connect an item to the record if you can connect it to the
>> manifestation itself? If an item is connected to a manifestation it
>> already is indirectly connected with the record/the description of
>> the manifestation.
> Ok, your are right.
>> Funnily, just today I had a related question about relating in RDF
>> journal to a library which at least holds one issue of this journal
>> and came up with the solution shown in. The initial problem was
>> we have URIs for journals in the German Union Catalogue of Serials
>> (Zeitschriftendatenbank, ZDB) and have URIs of libraries which own
>> some issues of this journal, without knowing which concrete issues
>> these are. As the issue as well as the item are unknown/have no URI
>> (yet), the result contains two blank nodes...
> Yes, it's better to have blank URIs than overhastyly introducing new

> URIs that must later be changed.

...especially, when URIs should better be minted by libraries in
namespaces controlled by them and not by the hbz. It's them who hold the
resource and know best about it's access conditions and availability


> Jakob
Received on Monday, 17 October 2011 12:40:36 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:27:44 UTC