UNIMARC in RDF project
- 1. The UNIMARC in RDF project: namespaces and linked data Mirna Willer, Gordon Dunsire, Predrag Perožid Presented at Session 222, IFLA WLIC 2013, 22 August 2013, Singapore
- 2. The project • Representation of UNIMARC structure and code lists in RDF (Resource Description Framework) • Funded for first year by Permanent UNIMARC Committee – Focus on UNIMARC Bibliographic format • Will take more than 1 year … – Further funding required …
- 3. The UNIMARC in RDF project aims to develop a representation of the UNIMARC format, including structures and code lists of controlled terminologies, in Resource Description Framework (RDF), the basis of the Semantic Web. It is funded for the first year by the Permanent UNIMARC Committee (PUC) which maintains the UNIMARC formats. The current focus is the UNIMARC Bibliographic format for descriptive records. The project will take more than one year to complete. Additional tasks include mapping the UNIMARC elements to reflect internal semantics and interoperate with other bibliographic formats. Notes for previous slide
- 4. Constructing the element set • “classes and attributes used to describe entities of interest” – W3C Library Linked Data Incubator Group • Represented as RDF classes and properties – In the Open Metadata Registry (OMR) – Uploaded from spread-sheets • Representation at finest level of granularity to avoid loss of information
- 5. The classes or types of entity of interest to UNIMARC, together with their attributes, are represented as a set of RDF classes and properties collectively known as an element set. The element set will be stored and managed using the Open Metadata Registry (OMR) which is also used for other IFLA bibliographic standards such as FRBR and ISBD. The data will be prepared offline using spread-sheets; the text of the UNIMARC format is not structured enough to allow direct import. The spread-sheets will then be uploaded into the OMR. UNIMARC elements are represented at the finest level of granularity so that all data in a UNIMARC record can be represented as linked data without any loss of content or contextual information. Notes for previous slide
- 6. tag tagCap ind1 ind1Cap ind2 ind2Cap sub subCap definition 210 PUBLICATION, DISTRIBUTION, ETC. # Not applicable / Earliest available publisher # Produced in multiple copies, usually published or publically distributed a Place of Publication, Distribution, etc. The town or other locality where the item is published or distributed or, in the case of a manuscript, written. 210 PUBLICATION, DISTRIBUTION, ETC. 0 Intervening publisher # Produced in multiple copies, usually published or publically distributed a Place of Publication, Distribution, etc. The town or other locality where the item is published or distributed or, in the case of a manuscript, written. 210 PUBLICATION, DISTRIBUTION, ETC. 1 Current or latest publisher # Produced in multiple copies, usually published or publically distributed a Place of Publication, Distribution, etc. The town or other locality where the item is published or distributed or, in the case of a manuscript, written. 210 PUBLICATION, DISTRIBUTION, ETC. # Not applicable / Earliest available publisher 1 Not published or publically distributed a Place of Publication, Distribution, etc. The town or other locality where the item is published or distributed or, in the case of a manuscript, written. 210 PUBLICATION, DISTRIBUTION, ETC. 0 Intervening publisher 1 Not published or publically distributed a Place of Publication, Distribution, etc. The town or other locality where the item is published or distributed or, in the case of a manuscript, written. 210 PUBLICATION, DISTRIBUTION, ETC. 1 Current or latest publisher 1 Not published or publically distributed a Place of Publication, Distribution, etc. The town or other locality where the item is published or distributed or, in the case of a manuscript, written.
- 7. tag tagCap ind1 ind1Cap ind2 ind2Cap sub subCap definition 210 PUBLICATION, DISTRIBUTION, ETC. # Not applicable / Earliest available publisher # Produced in multiple copies, usually published or publically distributed a Place of Publication, Distribution, etc. The town or other locality where the item is published or distributed or, in the case of a manuscript, written. 210 PUBLICATION, DISTRIBUTION, ETC. 0 Intervening publisher # Produced in multiple copies, usually published or publically distributed a Place of Publication, Distribution, etc. The town or other locality where the item is published or distributed or, in the case of a manuscript, written. 210 PUBLICATION, DISTRIBUTION, ETC. 1 Current or latest publisher # Produced in multiple copies, usually published or publically distributed a Place of Publication, Distribution, etc. The town or other locality where the item is published or distributed or, in the case of a manuscript, written. 210 PUBLICATION, DISTRIBUTION, ETC. # Not applicable / Earliest available publisher 1 Not published or publically distributed a Place of Publication, Distribution, etc. The town or other locality where the item is published or distributed or, in the case of a manuscript, written. 210 PUBLICATION, DISTRIBUTION, ETC. 0 Intervening publisher 1 Not published or publically distributed a Place of Publication, Distribution, etc. The town or other locality where the item is published or distributed or, in the case of a manuscript, written. 210 PUBLICATION, DISTRIBUTION, ETC. 1 Current or latest publisher 1 Not published or publically distributed a Place of Publication, Distribution, etc. The town or other locality where the item is published or distributed or, in the case of a manuscript, written. U21011a URI
- 8. tag tagCap ind1 ind1Cap ind2 ind2Cap sub subCap definition 210 PUBLICATION, DISTRIBUTION, ETC. # Not applicable / Earliest available publisher # Produced in multiple copies, usually published or publically distributed a Place of Publication, Distribution, etc. The town or other locality where the item is published or distributed or, in the case of a manuscript, written. 210 PUBLICATION, DISTRIBUTION, ETC. 0 Intervening publisher # Produced in multiple copies, usually published or publically distributed a Place of Publication, Distribution, etc. The town or other locality where the item is published or distributed or, in the case of a manuscript, written. 210 PUBLICATION, DISTRIBUTION, ETC. 1 Current or latest publisher # Produced in multiple copies, usually published or publically distributed a Place of Publication, Distribution, etc. The town or other locality where the item is published or distributed or, in the case of a manuscript, written. 210 PUBLICATION, DISTRIBUTION, ETC. # Not applicable / Earliest available publisher 1 Not published or publically distributed a Place of Publication, Distribution, etc. The town or other locality where the item is published or distributed or, in the case of a manuscript, written. 210 PUBLICATION, DISTRIBUTION, ETC. 0 Intervening publisher 1 Not published or publically distributed a Place of Publication, Distribution, etc. The town or other locality where the item is published or distributed or, in the case of a manuscript, written. 210 PUBLICATION, DISTRIBUTION, ETC. 1 Current or latest publisher 1 Not published or publically distributed a Place of Publication, Distribution, etc. The town or other locality where the item is published or distributed or, in the case of a manuscript, written. Place of publication … in Publication, distribution, etc. (Current or latest publisher) (Not published …) Label
- 9. One advantage of using a spread-sheet is that repetitive data can be easily copied. The finest level of granularity corresponds to the unique combination of UNIMARC tag, first and second indicators, and subfield. This requires a lot of data to be repeated while a single component varies. In this example there are 3 different first indicators and two different second indicators, resulting in 6 unique combinations. The URI of each combination is derived directly from the tag, indicators, and subfield. The label of each combination is derived from the subfield, tag, and indicator captions. The derivations are automated using spread-sheet formulas. However, the definition and scope note for each subfield has to be extracted manually from the UNIMARC text, which conflates definitions, scope notes, and usage notes. This requires significant intellectual intervention. Notes for previous slide
- 10. 200 1#$aBibliographica belgica $fCommission belge de bibliographie $f= Belgische Commissie voor bibliografie Semantic information embedded in content “= “ : Parallel U2001_f : First Statement of Responsibility ??? : Parallel First Statement of Responsibility
- 11. Another complication with the element set is that some UNIMARC semantic information is extended beyond the usual MARC structure and is effectively embedded in the content of subfields. In this example, the first subfield $f is represented with the URI U2001_f (reflecting the indicator values), and is the first statement of responsibility. But there is a second subfield f, which is not and cannot be a second “first” statement of responsibility. In fact, the equals (=) sign preceding the content is a generic indication that the subfield contains parallel information; in this case, a parallel first statement of responsibility. The project is investigating ways of representing this semantic information in the element set. Notes for previous slide
- 12. UNIMARC ISBD Property Label A Property Label U200__a Title proper = <> P1004 has title proper P1117 has title of individual work by same author P1137 has common title of title proper UNIMARC Alignment with ISBD Alignment is equal, broader, and narrower!
- 13. UNIMARC is aligned with ISBD, and a lot of alignment information is present in the text of UNIMARC. This information can be used to create RDF maps between the UNIMARC element set and the ISBD element set to support semantic interoperability between data from both sources. General alignments can indicate that the UNIMARC and ISBD elements are equivalent, or that one is broader in semantic scope than the other. Preliminary examination of the UNIMARC alignments with ISBD discloses cases where all three types of alignment seem to exist between the same pairs of elements, as shown in this example. This arises because the different semantics of some UNIMARC subfields are dependent on the presence of other subfields within the same tag. Again, the project will investigate how best to represent these cases in RDF. Notes for previous slide
- 14. Value vocabularies • “thesauri, code lists, term lists, classification schemes, subject heading lists, …” – W3C Library Linked Data Incubator Group • Often represented in RDF using Simple Knowledge Organization System (SKOS)
- 15. The UNIMARC Bibliographic format also specifies a large set of code lists to be used in coded information subfields. Such code lists can be represented as RDF value vocabularies, where each code in the list is assigned a URI. The attributes of the codes are then represented using properties from the Simple Knowledge Organization System (SKOS). Notes for previous slide
- 16. Attribute Character position Value Notes Type designator 0 c newspaper Frequency of issue l a daily Regularity 2 a regular 110 (CODED DATA FIELD: CONTINUING RESOURCES) $a (Continuing Resource Coded Data) Property URI = Subfield URI + Character position U110__a0 U110__a1 U110__a2
- 17. The method for assigning URIs to coded information subfields is an extension of the method for normal subfields. Coded subfields assign different types of code to different character positions with the subfield, so the character position, conventionally starting at 0, is added to the subfield URI to obtain a separate URI for each type of code. Notes for previous slide
- 18. unimarcb:U110__a1resource: 123 freq: a crtype: c unimarcb:U110__a0 reg: aunimarcb:U110__a2 “a” skos:notation skos:prefLabel “giornaliera”@it “diária”@pt “daily”@en Frequency map for Dublin Core, MARC 21, and RDA
- 19. This example shows how a coded information subfield is related or linked to its corresponding value vocabulary. The example is based on the daily newspaper example from the previous slide. The newspaper’s URI is 123, and is the subject of a set of data triples which use the RDF properties for the different types of code. The object of each triple is the URI of the particular code used. The code URI can then link internally to its attributes in the value vocabulary, which include the code as notation, and the preferred labels of the code in multiple languages. The frequency value vocabulary has been published by the project, together with translations of the term in Italian and Portuguese. The code URI can also link to external maps using alignments between specific codes or terms used in other value vocabularies, in this example taken from Dublin Core, MARC 21, and RDA: resource description and access. The data from the UNIMARC record can interoperate with data based on these other sources. Notes for previous slide
- 20. Thank you!