Hi - I'll note that the mapping decisions were made by our metadata services (then Cataloging) group, not by the tech folks making it all work, though we were all involved in the discussions.  One idea that came up was to do a, perhaps, lossy translation, but also stuff one triple with a text dump of the whole MARC record just in case we needed to grab some other element out we might need.  We didn't do that, but I still like the idea.  Ok, it was my idea.  ;)

D

-----Original Message-----
From: Code for Libraries [mailto:[log in to unmask]] On Behalf Of Esme Cowles
Sent: Monday, December 05, 2011 11:22 AM
To: [log in to unmask]
Subject: Re: [CODE4LIB] Models of MARC in RDF

I looked into this a little more closely, and it turns out it's a little more complicated than I remembered.  We built support for transforming to MODS using the MODS21slim2MODS.xsl stylesheet, but don't use that.  Instead, we use custom Java code to do the mapping.

I don't have a lot of public examples, but there's at least one public object which you can view the MARC from our OPAC:

http://roger.ucsd.edu/search/.b4827884/.b4827884/1,1,1,B/detlmarc~1234567&FF=&1,0,

The public display in our digital collections site:

http://libraries.ucsd.edu/ark:/20775/bb0648473d

The RDF for the MODS looks like:

        <mods:classification rdf:parseType="Resource">
            <mods:authority>local</mods:authority>
            <rdf:value>FVLP 222-1</rdf:value>
        </mods:classification>
        <mods:identifier rdf:parseType="Resource">
            <mods:type>ARK</mods:type>
            <rdf:value>http://libraries.ucsd.edu/ark:/20775/bb0648473d</rdf:value>
        </mods:identifier>
        <mods:name rdf:parseType="Resource">
            <mods:namePart>Brown, Victor W</mods:namePart>
            <mods:type>personal</mods:type>
        </mods:name>
        <mods:name rdf:parseType="Resource">
            <mods:namePart>Amateur Film Club of San Diego</mods:namePart>
            <mods:type>corporate</mods:type>
        </mods:name>
        <mods:originInfo rdf:parseType="Resource">
            <mods:dateCreated>[196-]</mods:dateCreated>
        </mods:originInfo>
        <mods:originInfo rdf:parseType="Resource">
            <mods:dateIssued>2005</mods:dateIssued>
            <mods:publisher>Film and Video Library, University of California, San Diego, La Jolla, CA 92093-0175 http://orpheus.ucsd.edu/fvl/FVLPAGE.HTM</mods:publisher>
        </mods:originInfo>
        <mods:physicalDescription rdf:parseType="Resource">
            <mods:digitalOrigin>reformatted digital</mods:digitalOrigin>
            <mods:note>16mm; 1 film reel (25 min.) :; sd., col. ;</mods:note>
        </mods:physicalDescription>
        <mods:subject rdf:parseType="Resource">
            <mods:authority>lcsh</mods:authority>
            <mods:topic>Ranching</mods:topic>
        </mods:subject>

etc.


There is definitely some loss in the conversion process -- I don't know enough about the MARC leader and control fields to know if they are captured in the MODS and/or RDF in some way.  But there are quite a few local and note fields that aren't present in the RDF.  Other fields (e.g. 300 and 505) are mapped to MODS, but not displayed in our access system (though they are indexed for searching).

I agree it's hard to quantify lossy-ness.  Counting fields or characters would be the most objective, but has obvious problems with control characters sometimes containing a lot of information, and then the relative importance of different fields to the overall description.  There are other issues too -- some fields in this record weren't migrated because they duplicated collection-wide values, which are formulated slightly differently from the MARC record.  Some fields weren't migrated because they concern the physical object, and therefore don't really apply to the digital object.  So that really seems like a morass to me.

-Esme
--
Esme Cowles <[log in to unmask]>

"Necessity is the plea for every infringement of human freedom. It is the  argument of tyrants; it is the creed of slaves." -- William Pitt, 1783

On 12/3/2011, at 10:35 AM, Karen Coyle wrote:

> Esme, let me second Owen's enthusiasm for more detail if you can 
> supply it. I think we also need to start putting these efforts along a 
> "loss" continuum - MODS is already lossy vis-a-vis MARC, and my guess 
> is that some of the other MARC->RDF transforms don't include all of 
> the warts and wrinkles of MARC. LC's new bibliographic framework 
> document sets as a goal to bring along ALL of MARC (a decision that I 
> think isn't obvious, as we have already discussed here). If we say we 
> are going from MARC to RDF, how much is actually captured in the 
> transformed data set? (Yes, that's going to be hard to quantify.)
> 
> kc
> 
> Quoting Esme Cowles <[log in to unmask]>:
> 
>> Owen-
>> 
>> Another strategy for capturing MARC data in RDF is to convert it to MODS (we do this using the LoC MARC to MODS stylesheet: http://www.loc.gov/standards/marcxml/xslt/MARC21slim2MODS.xsl).  From there, it's pretty easy to incorporate into RDF.  There are some issues to be aware of, such as how to map the MODS XML names to predicates and how to handle elements that can appear in multiple places in the hierarchy.
>> 
>> -Esme
>> --
>> Esme Cowles <[log in to unmask]>
>> 
>> "Necessity is the plea for every infringement of human freedom. It is 
>> the argument of tyrants; it is the creed of slaves." -- William Pitt, 
>> 1783
>> 
>> On 11/28/2011, at 8:25 AM, Owen Stephens wrote:
>> 
>>> It would be great to start collecting transforms together - just a 
>>> quick brain dump of some I'm aware of
>>> 
>>> MARC21 transformations
>>> Cambridge University Library - http://data.lib.cam.ac.uk - 
>>> transformation made available (in code) from same site Open 
>>> University - http://data.open.ac.uk - specific transform for 
>>> materials related to teaching, code available at 
>>> http://code.google.com/p/luceroproject/source/browse/trunk%20lucerop
>>> roject/OULinkedData/src/uk/ac/open/kmi/lucero/rdfextractor/RDFExtrac
>>> tor.java (MARC transform is in libraryRDFExtraction method) COPAC - 
>>> small set of records from the COPAC Union catalogue - data and 
>>> transform not yet published Podes Projekt - LinkedAuthors - 
>>> documentation at 
>>> http://bibpode.no/linkedauthors/doc/Pode-LinkedAuthors-Documentation
>>> .pdf - 2 stage transformation firstly from MARC to FRBRized version 
>>> of data, then from FRBRized data to RDF. These linked from 
>>> documentation Podes Project - LinkedNonFiction - documentation at 
>>> http://bibpode.no/linkednonfiction/doc/Pode-LinkedNonFiction-Documen
>>> tation.pdf - MARC data transformed using xslt 
>>> https://github.com/pode/LinkedNonFiction/blob/master/marcslim2n3.xsl
>>> 
>>> British Library British National Bibliography - 
>>> http://www.bl.uk/bibliographic/datafree.html - data model 
>>> documented, but no code available Libris.se - some notes in various 
>>> presentations/blogposts (e.g. 
>>> http://dc2008.de/wp-content/uploads/2008/09/malmsten.pdf) but can't 
>>> find explicit transformation Hungarian National library - 
>>> http://thedatahub.org/dataset/hungarian-national-library-catalog and 
>>> http://nektar.oszk.hu/wiki/Semantic_web#Implementation - some 
>>> information on ontologies used but no code or explicit 
>>> transformation (not 100% sure this is from MARC) Talis - implemented 
>>> in several live catalogues including 
>>> http://catalogue.library.manchester.ac.uk/  - no documentation or 
>>> code afaik although some notes in
>>> 
>>> MAB transformation
>>> HBZ - some of the transformation documented at https://wiki1.hbz-nrw.de/display/SEM/Converting+the+Open+Data+from+the+hbz+to+BIBO, don't think any code published?
>>> 
>>> Would be really helpful if more projects published their 
>>> transformations (or someone told me where to look!)
>>> 
>>> Owen
>>> 
>>> Owen Stephens
>>> Owen Stephens Consulting
>>> Web: http://www.ostephens.com
>>> Email: [log in to unmask]
>>> Telephone: 0121 288 6936
>>> 
>>> On 26 Nov 2011, at 15:58, Karen Coyle wrote:
>>> 
>>>> A few of the code4lib talk proposals mention projects that have or will transform MARC records into RDF. If any of you have documentation and/or examples of this, I would be very interested to see them, even if they are "under construction."
>>>> 
>>>> Thanks,
>>>> kc
>>>> 
>>>> --
>>>> Karen Coyle
>>>> [log in to unmask] http://kcoyle.net
>>>> ph: 1-510-540-7596
>>>> m: 1-510-435-8234
>>>> skype: kcoylenet
>> 
> 
> 
> 
> --
> Karen Coyle
> [log in to unmask] http://kcoyle.net
> ph: 1-510-540-7596
> m: 1-510-435-8234
> skype: kcoylenet