Linking Library of Congress Subject Headings Owen Stephens 14th July 2011 @ostephens http://www.meanboyfriend.com/overdue_ideasThursday, 14 July 2011
LNKNLCSH @ostephensThursday, 14 July 2011This is the lightning versionI should precursor this talk by saying I’m really pleased that the LoC have invested inexperimenting with Linked Data representations of aspects of their data. Anything in this talkisn’t a criticism of this, but about the issues we encountered using aspects of the data. It’spossible that some or all of these problems may have been down to my lack of understandingof LCSH and Linked Data :)
Library chopsThursday, 14 July 2011I’m a librarian - by nature and qualiﬁcation :) - see http://www.meanboyfriend.com/overdue_ideas/2010/11/library-routes/Been working on the cusp between libraries and IT since 1995. Spending early part of mycareer in small libraries means I have worked in just about every area of library front of houseand back office. However although I’ve catalogued books, and have more than a passingfamiliarly with MARC, I’m not a cataloguer, and not an expert on LCSH
Linked Data chopsThursday, 14 July 2011I’ve been trying to understand the Semantic Web/Linked Data for several years :) Myunderstanding has been accelerated over the last couple of years by involvement in severalprojects in the Linked Data space. Speciﬁcally the Lucero and CORE projects at the OpenUniversity
Thursday, 14 July 2011Expressing similarity between published papers in UK research repositoriesHarvest metadata and full-text (50k papers from 143 UK repos so far)Text mine for relationshipsExpose ‘similarity’ measure as RDF triples using MuSIM Ontology (originally developed forMusic, but equally applicable)For more information http://core-project.kmi.open.ac.uk
Exposing RDFThursday, 14 July 2011Three ʻproductsʼCORE Portal - search or SPARQL metadata for harvested papersCORE Mobile – Android application to search & navigate across related papers & downloading articlesCORE Plugin - Designed to integrate into existing repository interface to link to ʻrelated papersʼ in other repos, based on CORE ʻsimilarityʼFor more information http://core-project.kmi.open.ac.ukSPARQL Endpoint at http://core.kmi.open.ac.uk:8081/COREWeb/squeryHow we express data in RDF: http://core-project.kmi.open.ac.uk/node/13
LuceroThursday, 14 July 2011For more information see http://lucero-project.infoData and SPARQL Endpoint available via http://data.open.ac.ukLucero published variety of data from the Open University as linked open data - admin data(buildings), course data (course catalogue, OERs), research data and data about bibliographicresources - including materials in the library (focussed on materials related to coursematerials - around 30k catalogue records)
LCSHThursday, 14 July 2011Lots been written about LCSH, it’s structure, whether it should be replaced. I don’t want to spend too much time on this today but it may come up in placesHowever it is probably worth recapping my understanding (if only to let those more knowledgeable correct it)Key aspect in the context of this talk is that LCSH is primarily a pre-coordinated system - that is facets of subject headings are pre-combined into a single,multi-faceted heading.Although....“LCSH itself requires some degree of post-coordination of the pre-coordinated strings to bring out speciﬁc topics of works.” (http://www.loc.gov/catdir/cpso/pre_vs_post.pdf)In fact the way that LCSH is structured in MARC records, and the way that indexes can be built on this in library management systems means thatI’m going to focus on ‘Topical’ subject headings (confusingly to me, LCSH can also cover Name, Title and Geographic headings)Topical Terms can represent “a concrete object, animal, etc.; a category of people, animals, or objects; a more abstract concept, belief, process, orphenomenon; an institution, etc.” (http://www.tulane.edu/~techserv/lcsh%20introd.html)Topical LC Subject Headings are built by combining ‘Topical Terms’ with qualiﬁers (‘subdivisions’) which allow you to contextualise the term.The types of subdivision available are:General (a high level general qualiﬁer - e.g. ‘History’)Chronological (period of time - e.g. ‘20th Century’)Geographic (place - e.g. ‘Great Britain’)Form (the type/genre of material - e.g. ‘Dictionary’)There are large number of rules that express how these subdivisions can be used in conjunction with Topical Terms, and the order in which they should beexpressed. Not all combinations are valid - for example only certain General subdivisions may be further subdivided Geographically. The rules are not alwaysblack and white - they have ‘examples’ lists which you can use to inform you if it might be valid in a given situation.Perhaps suffice to say that a document called ‘BASIC SUBJECT CATALOGING USING LCSH: Trainee’s Manual’ is 382 pages long.Subject heading strings can be valid (i.e. constructed according to rules/patterns) while not being ‘Authorized’ - in this context and Authorized Heading is “Apreferred subject term as decided and established by the Library of Congress by means of an authority record.” (Thanks to Tom Meehan for this deﬁnition)
Thursday, 14 July 2011Thanks to work of Ed Summers and others, the Library of Congress have a Linked Datarepresentation of LCSH in SKOS. However, this only covers ‘Authorized’ LCSH - presumablybecause only those LCSH with an Authority record have an identiﬁer within LoC systems? (I’mspeculating)
Thursday, 14 July 2011This is a catalogue record from the OU - the two strings listed as ‘Subjects’ are LCSH (forcataloguers amongst you MARC 650s)Can see the linked data representation at http://data.open.ac.uk/page/library/289148
General Subdivision Science--Study and Teaching--Research Topical General Term SubdivisionThursday, 14 July 2011This is made up of a Topical Term - Science and two general subdivisions ‘Study andTeaching’ and ‘Research’
Science--Study and Teaching--Research id.loc.gov ?Thursday, 14 July 2011This is (afaik - I trust the cataloguers) a valid LCSH ... however it is not authorized ... and sodoes not have a URI on id.loc.gov
Science--Study and Teaching--Research http://id.loc.gov/authorities/sh85118587#conceptThursday, 14 July 2011“Science--Study and Teaching”, however, is an authorized heading
Science--Study and Teaching--Research http://id.loc.gov/authorities/sh85118553#concept N.B. This is URI for Science as Topical Term not http://id.loc.gov/authorities/sh00007934#concept which is URI for Science as a General SubdivisionThursday, 14 July 2011As is “Science”
Science--Study and Teaching--Researchhttp://id.loc.gov/authorities/sh2001008697#conceptThursday, 14 July 2011Also “Study and Teaching” (as a topical subdivision) is an authorized heading
Science--Study and Teaching--Researchhttp://id.loc.gov/authorities/sh2002006576#concept N.B. This is URI for Research as General Subdivision not http://id.loc.gov/authorities/sh85113021#concept which is URI for Research as a Topical TermThursday, 14 July 2011Also “Research” (as a topical subdivision) is an authorized heading
More links pleaseThursday, 14 July 2011If we only used id.loc.gov URIs where we had an authorised LCSH, we would end up with onlya small number of links. Some URIs in id.loc.gov would never be used in this way as they onlyrepresent subdivisions - never valid by themselves.Therefor decided to check a variety of combinations against id.loc.gov
Science--Study and Teaching--Research Science--Study and teaching http://id.loc.gov/authorities/ sh85118587#concept http://id.loc.gov/authorities/ Science sh85118553#concept http://id.loc.gov/authorities/ Study and Teaching sh2001008697#concept http://id.loc.gov/authorities/ Research sh2002006576#conceptScience--Study and Teaching-- http://data.open.ac.uk/page/topic/ library/science-- Research study_and_teaching--researchThursday, 14 July 2011
MADS? http://www.loc.gov/standards/mads/rdf/Thursday, 14 July 2011As far as I can see MADS (apart from looking complex) models the Authority - not theheading - this doesn’t solve the problem we saw here!That is MADS would solve the problem only for Authorized headings (which it does representas component parts - which I think addresses the issues raised by Karen Coyle at http://kcoyle.blogspot.com/2009/05/lcsh-as-linked-data-beyond-dash-dash.html)Happy to be corrected...
A different approach? bibo:authorList ( <http://examples.net/contributors/2> <http://examples.net/contributors/1>) lcsh:headingList ( <http://id.loc.gov/authorities/ sh85118553#concept> <http://id.loc.gov/authorities/ sh2001008697#concept> <http://id.loc.gov/authorities/ sh2002006576#concept>)Thursday, 14 July 2011If we could use rdfs:list to represent the pre-coordinated string of headings - then wouldn’tcare about whether ‘authorized’ or not, and would have all the individual headings there aswell (bibo lists authors individual and as a list)Again copying BIBO which has each author as a dc:author as well, could represent each partof the subject string as a separate dc:subject.In a MADS world there would be advantage to expressing full authorized heading as well (forrelationships derived in MADS) although there is still the question of expressing ‘authorizedfragments’ which seems to me would also be useful with MADS for the same reasonsThis feels like a simple approach that would at least allow us to capture the component partsof subject string (and personally I’m not sure we ought to go further than this? do we needto? why?). My feeling is lots of the work goes into representing the ‘Authority ﬁle’ as opposedto how subject headings are used in the real world ... is this fair?
Details: http://discovery.ac.uk/developers/ competition/ Datasets: http://ckan.net/group/ ukdiscovery Ask Questions: http://getthedata.org or #discodevThursday, 14 July 2011Finally just an advert - if you are interested in open data in the library/archive/museumspace please consider entering this competition :) - really show the value of this stuff!