View
213
Download
1
Category
Preview:
Citation preview
http://www.ukoln.ac.uk/
Metadata Schema Registries in the Partially Semantic Web: the CORES experience
Rachel Heery, Pete Johnston, UKOLN, University of Bath
András Micsik, Csaba Fülöp, MTA SZTAKI, Budapest
DC-2003, Seattle, Washington, USA
28 September – 2 October, 2003
http://www.ukoln.ac.uk/
The CORES experience
• The CORES project
• The CORES registry– The registry & the schema creation tool– The registry data model
• The CORES registry in practice
http://www.ukoln.ac.uk/
The CORES project
http://www.ukoln.ac.uk/
The CORES project
• Funded within European Commission FP5 IST Programme
• Partners– PricewaterhouseCoopers Luxembourg– Fraunhofer Gesellschaft– UKOLN, University of Bath– MTA SZTAKI, Budapest
• To encourage the sharing of metadata semantics– Standards Interoperability Forum– Metadata Schema Registry
http://www.ukoln.ac.uk/
The CORES registry
http://www.ukoln.ac.uk/
What is the CORESmetadata schema registry?
• Software application that provides access to information on metadata element sets, their constituent elements, and their use
• Primarily to support disclosure and discovery, but also reuse
• Information provided to registry in the form of machine-readable schemas
• Interfaces for human readers and software applications
What is the CORESmetadata schema registry?
CORES Registry
Schema
Schema
Schema
Schema
http://www.ukoln.ac.uk/
The registry data model
• Builds on earlier work in DESIRE, SCHEMAS projects
• Dublin Core "grammatical principles"– Element refinement– Encoding Scheme– Resource – Property (DC Element or DC
Element refinement) – Value– Value may be associated with Encoding
Scheme
http://www.ukoln.ac.uk/
The registry data model
• The concept of the "application profile"– Metadata elements are defined/managed as
members of metadata element sets– Implementers “optimise” their use of metadata
elements• May constrain usage of elements in context• May narrow "standard" element semantics• May draw elements from multiple element sets
AgencyElement
Set1m
Encoding Scheme
1
m
m m(Controlled
Vocab)Value0 m
m
m
1
m ElementUsage
1m AppProfile
1
m
Element
1
mm
1
http://www.ukoln.ac.uk/
The registry software
• Software development by– András Micsik & Csaba Fülöp (MTA SZTAKI)
• Enhancing MEG Registry project software– Dave Beckett & Damian Steer (ILRT)
• Schema creation tool– Java Swing, Jena RDF toolkit– Forms-based authoring– Save/reload as RDF/XML– "Use" existing Elements by "drag-and-drop" from
results of search of registry database– Submit data to registry (HTTP POST)
CORES Schema Creation Tool
Search registry for
“title”Elements
Drag Elementto create
Element Usage in App Profile
http://www.ukoln.ac.uk/
The registry software
• Registry server– Perl/CGI, Redland RDF application
framework– Upload/publication API (HTTP POST)– Simple query API (HTTP GET)
• Very basic, designed to support Schema Creation Tool
– HTML interface • Browse, query, annotate, administer
CORES Registry HTML interface
http://www.ukoln.ac.uk/
The CORES registry in practice
http://www.ukoln.ac.uk/
The CORES registry, August 2003
• Registered– 11 Element Sets (152 Elements)– 86 Encoding Schemes– 9 Application Profiles (254 Element Usages)
• Mostly DC-based
• Sources– Schemas created using CORES tool– Existing RDFS data published on Web
http://www.ukoln.ac.uk/
Using existing RDFS data
• CORES tools built on registry data model– uses RDFS– but also application-specific RDF vocabulary
• Schema Creation Tool– does not load "pure" RDFS documents
• Registry server– does read/index RDFS data– but requires supplementary data to describe
application-specific attributes, relations
http://www.ukoln.ac.uk/
Using existing RDFS data
• Use of existing RDFS/RDFVDL data is possible but not straightforward using CORES registry tools
http://www.ukoln.ac.uk/
Providing access to source Schemas
• Schema creation tool – creates descriptions of entities in model– stores descriptions as Schema (RDF/XML)– does not assume one-to-one relations
• between Element Set and Schema• between Application Profile and Schema
– does not create description of Schema itself
• Registry server– has no metadata about Schema– does not maintain record of source of data
http://www.ukoln.ac.uk/
Providing access to source Schemas
• Registry provides access to descriptions of resources submitted, but does not provide direct access to source Schemas
• Suggest adding Schema to model as entity, amend tools to generate/use rdfs:isDefinedBy statements
http://www.ukoln.ac.uk/
Availability of CORES registry server(s)
• CORES as fixed-term project– Unable to guarantee long-term availability of
current CORES registry service– Schema owners reluctant to invest effort in
creating data for registry
• Distinction between– Schemas created using the CORES tools– Service currently provided by MTA SZTAKI
• Guaranteed by "Persistence Policy" till mid 2004
– The CORES registry software• Available from Sourceforge
http://www.ukoln.ac.uk/
Availability of CORES registry server(s)
• Continued availability of Schemas is independent of availability of SZTAKI registry
• Other service provider can provide registry server using CORES registry software– Re-index existing Schemas from Web
• Data also available to other RDF/RDFS applications– But N.B. Schema Creation Tool uses CORES
RDF vocabulary
http://www.ukoln.ac.uk/
Shared models for metadata
• Registry based on DC metadata model• But metadata vocabularies may have own
(unrelated) metadata model– "Element" != CORES Element/RDF Property
• Some metadata vocabularies defined primarily in terms of XML tree structure – XML element != CORES Element/RDF Property
• Even where vocabulary has RDF expression, additional effort – e.g. RDFS Class != CORES Encoding Scheme
http://www.ukoln.ac.uk/
Shared models for metadata
• May be possible to map from "native" model to CORES model, but – requires element-by-element analysis– different process for every vocabulary – arguably, result of limited value to implementer
working only with "native" model
• Conversely, registry can not "predict" structural conventions of arbitrary XML encodings– Application Profile metadata includes optional
pointer to XML Schema
http://www.ukoln.ac.uk/
Integrity and trust
• Basis of Application Profile is reuse of existing Elements, Encoding Schemes– resources defined/published by others in a
global space
• Expectation that the URI will continue to denote what it denotes at the time of reuse
• Requires level of trust – in source/owner of URI– in mediating service (registry) that exposes
metadata about that resource
http://www.ukoln.ac.uk/
Conclusions
• Registry data model is simplification of complex reality– Good "fit" for DC Application Profiles– More problematic where models diverge
• Application-specific RDF vocabulary does limit interoperability– Review in light of recent RDF specs, OWL
• Trust issues require work• Shared model is critical, especially where
reuse encouraged
http://www.ukoln.ac.uk/
Acknowledgements
• UKOLN is funded by Resource: the Council for Museums, Archives and Libraries, the Joint Information Systems Committee (JISC) of the UK higher and further education funding councils, as well as by project funding from the JISC and the European Union. UKOLN also receives support from the University of Bath where it is based.
• http://www.ukoln.ac.uk/
http://www.ukoln.ac.uk/
Metadata Schema Registries in the Partially Semantic Web: the CORES experience
Rachel Heery, Pete Johnston, UKOLN, University of Bath
András Micsik, Csaba Fülöp, MTA SZTAKI, Budapest
DC-2003, Seattle, Washington, USA
28 September – 2 October, 2003
Recommended