The Vocabulary Mapping Framework matrix: background ... and application
description
Transcript of The Vocabulary Mapping Framework matrix: background ... and application
The Vocabulary Mapping Framework matrix: background ... and application
Gordon DunsirePresented to the Vocabulary Mapping
Framework seminar
9 Nov 2009, British Library, London
Vocabulary Mapping Framework
Funded by UK’s Joint Information Systems Committee (JISC)Only first stage funded
Major expansion of the RDA/ONIX frameworkTo create a tool to support the automated
mapping of vocabularies from metadata standards of use to the JISC community
Research, teaching, learningProject conducted during second half of 2009
A starting point: RDA outreach
RDA outreach to other communitiesRDA/ONIX frameworkDCMI (Dublin Core Metadata Initiative) RDA Task
GroupMembers from Dublin Core, IEEE-Learning Object
Metadata, RDA, and W3C communities
Expressing RDA element set and value vocabularies in Resource Description Framework (RDF)
RDA alignment
RDA alignment with recent metadata models developed by IFLA (International Federation of Library Associations and Institutions)Functional Requirements for Bibliographic
Records (FRBR)Functional Requirements for Authority Data
(FRAD)Statement of International Cataloguing Principles
Stimulated IFLA project to develop RDF representation of FRBRExpected end of 2009
IFLA 2009
Further discussion at IFLA’s World Library and Information Congress 2009 (Milan, Italy)Study group to consider RDF/XML representation
of International Standard Bibliographic Description (ISBD)
Model underpinning many national cataloguing schema, including MARC21
Task group to consider general support for RDF/XML namespaces for IFLA “standards”
Consolidation of “FR” family of models, including Functional Requirements for Subject Authority Data (FRSAD) when finalised (2010+)
Linked data (2009, mostly)
Increasing presence of “expert” metadata in the linked data pool
Library of Congress Subject Headings (LCSH)With relators to Rameau (French subject heading
scheme) terms
Top-level Dewey Decimal Classification (DDC) notations and captions (1000+)In 9 languages
Top-level Universal Decimal Classification (UDC) notations and captions real soon now
Opportunities and possibilities
If metadata schema (MARC21, UNIMARC, RDA, ISBD) in RDFThen easier to parse instance data (catalogue
records) into RDF
If very large quantities of legacy instance data available in RDFThen latent associations (relationships) can be
identified using statistical inferencingE.g. Mapping of DDC notations to LCSH (WebDewey)
If critical mass of rich (diverse) RDF triplesThen utility of Semantic Web increases
Threats and vulnerabilities
RedundancySame relationship/property expressed multiple
times in multiple ways makes inferencing more complex
GapsRelationship/property not expressed at all so
classes/entities remain isolated
Authority and authenticityWho says this? Professional expert, end-user,
machine?
Currency and versionWhen was this said? Is it deprecated?
VMF requirements
Scalable and extensible to accommodate new and changing vocabularies
Flexible to allow engagement by different communities in various stages of vocabulary development and mapping
Non-prescriptive to encourage uptakeAnd allow use beyond VMF (and RDF)
environment
... Goto: How the VMF matrix works / Godfrey Rust
Applications
Metadata cross-walksBetween different vocabularies
E.g. Publisher metadata (ONIX) and library metadata (RDA)
Mapping of local, bespoke metadata schemesFrom local scheme to global frameworkLocal metadata often specialised, specific, and
unique
Application: Complex term mapping
Different domains may have very different views on resource categorization and identificationE.g. Institutional repositories may distinguish draft
versions of scholarly articles while publishers need only “see” the final version
Conversely, publishers may distinguish the components of a monograph for rights assignment (text, illustrations, binding, etc.), while libraries want to treat it as a single resource
Good relator mappings can improve metadata exchange across differing granularities
Translations
Translation of the VMF matrix is unnecessaryIntended for machine use only
Several ways of handling translations of mapped vocabulariesDeclare translations within the term (@en, @de,
etc.)Same term URI already mapped to VMF
Declare translations as a separate vocabularyThen map term URIs to original vocabulary
Original URIs already mapped to VMF
Or map term URIs directly to VMF
Modes of use
RDF properties/relators vs RolesRole (contextual): writerProperty/Relator pair: writes; isWrittenBy
Formal vs Remote vs InformalTerm mapping contained within “official” version
of the VMFTime Authenticity Reliability Utility
Term mapped to term within official versionTime Authenticity ? Reliability ? Utility
Copy of VMFTime Authenticity ? Reliability Utility ?
Concept family
Accommodates terms for roles, bi-directional relator pairs, uni-directional relators (properties), classes and attributes
FRBR class “Choreography”vmf:ChoreographedDance
RDA role “choreographer”vmf:ChoreographedDance_DanceChoreographer
RDA/ONIX attribute “language”vmf:LexicalWork
DDEX role “Author”Vmf:LexicalWork_Writer
Roles vs relators
Vocabularies of role terms are popularDDEX, MARC, ONIX, RDA ...
Role definitions can be extracted from VMFHub (vmf) and spoke (mapped)From a local copy
May aid development of role vocabularies by other communitiesWith possible subsequent incorporation in VMF
(VMF as guidance on good (shared) practice)
Identification
VMF namespace URI for every mapped vocabulary termLinked to published (scheme) URI for termScheme URI to be used as external referent
VMF is a black box
If term not mapped within VMFAdd to VMF
Forces review of internal VMF mappingsIf no scheme URI, use VMF URI as referent?
Map to term already mapped within VMFowl:equivalentClass; owl:equivalentPropertyNature of equivalence?
Sticking to VMF
VMF
Semantic (fractal) soup?
Thank you
Questions?