Re: Antw: Re: Citation Cluster

Hi Adrian,
> These questions seem to be related. I think this depends on the use
> case. If you want to model "legacy references" in RDF you probably
> should identify references as is done in your illustration so you can
> associate them with the original string. This comes handy when you have
> to compare two references which are different strings but might identify
> the same resource and could possibly be linked. In future scholarly
> texts, though, this shouldn't be necessary at all because linking should
> exclusively happen by using URIs and not strings.
Agreed, but I think it will need some time until we reach this future :-)

> So - to understand what this is all about - what is the use case you
> are discussing? Is it about digitizing works, doing OCR and parsing
> citations and references to triplify them and link them to bibliographic
> entries in library services etc.?
Yes.
> Or is it about creating research
> production tools where links to other resources can be establuished
> easily? As you are talking about books and pages in the visualization
> you are probably talking about the first use case. Thus, in my opinion,
> you should model a bibliographic reference as an own class and associate
> with it the actual string from the respective bibliography.
>
And yes. As we are curating a cluster of use cases, both (and more) are 
relevant scenarios that affect the proper representation of citations as 
linked data.
>> Another question might be, if we want to repeat the bibliographic
>> information for the reference, if we can directly link the
> bibliographic
>> resource that is identified by the reference.
> I think if "we can directly link the bibliographic resource that is
> identified by the reference" means: We have a URI (DOI, URN etc.) for
> this resource, then we probably can discard the bibliographic
> information. (On the other side Ben wrote about "dois that point to the
> wrong article" which would justify keeping the actual reference string.)
> If it means: we can link to a bibliographic resource only based on the
> provided string then we should keep the string.
Probably, especially in the "future" scenario, were the link is directly 
established by the author and the printed reference list is probably 
generated automatically from the linked resource and its description. So 
a good practice would be to keep the original string information only 
for the "legacy" resources, where they were used to establish the link. 
So to speak data from the dark "pre linked data" age of mankind, still 
ongoing ;-)
> One thing that doesn't seem right to me in the visualization: the
> dc:creator-relation "Bell, Stewart" is associated with the reference. If
> a creator is associated with the reference it shouldn't be the creator
> of the referenced work but the creator of the referencing work (in this
> case: Bibliographic Resource :book1).
Oops, you are totally right! That shows that you really, really have to 
be careful, when you reuse element vocabularies. I will replace it or 
think about a different solution.

Cheers,

Kai

-- 
=============================================
Kai Eckert
KR&  KM Research Group
Universität Mannheim
B6, 23-29; Building B; Room B 1.15
D-68159 Mannheim
Tel.:  +49 621 181 2332
Fax:   +49 621 181 2682
WWW:   http://ki.informatik.uni-mannheim.de
---------------------------------------------

Received on Friday, 10 December 2010 12:10:40 UTC