Terminology mapping for subject cross-browsing in distributed information environments

1
Terminology mapping for subject cross-browsing in
distributed information environments

• Libo Si
• PhD student in the Department of Information
  Science, Loughborough University

2
Background

• Users have to face different information
  resources using different schemes.
• Library portal systems, such as MetaLib, SirSi
  Room.
• These provide a single access point.

3
Background

• Keyword cross-searching
• Mapping different metadata schemes.
• Make them interoperable.
• Subject cross-browsing
• Integrate different KOSs together into a
  hierarchical tree.
• Issues
• Interoperability between different knowledge
  organisation systems
• Interoperability between metadata standards

4
My Research

• Aim
• To develop methods to facilitate both subject
  cross-browsing and cross-searching for library
  portal systems.
• Objectives
• To investigate different methods to develop
  cross-search service in a library portal product
  
• To investigate different methods to make
  different metadata standards interoperable
• To investigate different methods to make
  different knowledge organisation systems
  interoperable
• To indicate some trends to establish ontologies
  to facilitate both cross-searching and
  cross-browsing by subject for the development of
  library portal systems.

5
Methodology

• Case study HILT, Renardus, MetaNet, ABC
  Ontology, OpenCyc Ontology, ePrint UK, and UMLS.
• Investigate different methods used by these
  projects to facilitate subject cross-browsing and
  cross-searching service.

6
Methods to cross-search (1)
Federated Search (Sadeh 2006)
7
Methods to cross-search (1)

• A cross-search service can create and maintain
  their own repository of resource metadata (Sadeh
  2004).
• Issues
• Loss of data value
• Cannot capture rich knowledge organisation
  systems used by different online databases due to
  the lack of methods to reuse different metadata
  schemes and controlled vocabularies (Hughs and
  Kamat 2005).

8
Methods to cross-search (2)

• An alternative is
• In the semantic web community, the construction
  of ontologies to maximise the use of both subject
  classification systems and metadata schemes
  across different collections is possible.
• Each participating resource providers can offer
  metadata and classification systems to any
  cross-search service.

9
Mapping semantics of different metadata standards

• Derivation
• Application profile
• Crosswalk (one-to-one, and switch)
• Metadata registry
• Data reuse and integration (RDF)
• Aggregation.
• - Chen and Zheng (2006)

10
Derivation

• One metadata scheme can be developed based on the
  principle and structure of an existing one (Chan
  and Zeng 2006a).
• Ex. TEI Lite is derived from the full Text
  Encoding Initiative (TEI).

11
Application profile

• An application profile can be defined by
  combining a selected range of metadata elements
  from different metadata schemes for some
  application-specific purpose (Heery and Patel
  2004).

12
Project using Application Profile

• Five namespaces used by Renardus application
  profile
• http//renardus.sub.unigoettingen.de/renap/renap.h
  tml
• Renardus Metadata Element Set (rmes),
• Renardus Metadata Element Set Qualifiers (rmesq),
  
• Dublin Core Metadata Element Set, version 1.1 (dc
  1.1),
• Dublin Core Metadata Element Set Qualifiers
  (dcterms),
• DCMI Type Vocabulary (dcmitype).

13
Crosswalk

• A crosswalk is a specification for mapping one
  metadata standard to another (St. Pierre and
  LaPlant 1998).
• One-to-one
• Many-to-many (switch scheme)

14
Metadata scheme registry

• A metadata registry refers to an application that
  provides services based on information about
  'metadata terms' and about related resources
  (Johnston 2005).
• Ex the CORES registry lists more than 40
  metadata schemes, and supports searching and
  browsing by metadata scheme developer,
  maintenance agency, element sets, elements,
  encoding schemes, application profiles and
  element usages.
• (http//www.cores-eu.net/registry/)

15
Data reuse and integration

• This refers to describing information objects by
  using different elements from different metadata
  schemes or application profiles (Chan and Zeng
  2006b).
• The Resource Description Framework (RDF) provides
  a basic platform for integrating different
  metadata schemes to describe web resources (Heery
  and Patel 2004).
• RDF can facilitate the use of different
  application profiles.

16
An RDF example
lt?xml version"1.0" ?gt - ltrdfRDF
xmlnsrdf"http//www.w3.org/1999/02/22-rdf-syntax
-ns" xmlnsdchttp//purl.org/dc/elements/1.0/
xmlnsbc"http//www.schemas-forum.org/registry/sc
hemas/BIBLINK/1.0/bc-ap"gt - ltrdfDescription
about"urnisbn0-89887-113-1"gt  
ltdctitlegtPatrologia Latina Databaselt/dctitlegt
  ltdccreatorgtJacques Paul Mignelt/dccreatorgt
  ltdcdategt1993lt/dcdategt  
ltdclanguagegtlalt/dclanguagegt    ltbcextentgt2
computer laser optical disks 4 3/4
inlt/bcextentgt    ltbcsystemRequirementsgtMultim
edia PC 486x or higher, 8mb memory, CD-ROM drive,
sound card, SVGA 256-colour monitor, Windows 95
or Windows 3.1lt/bcsystemRequirementsgt  
ltdcsubject rdfvalue"Christian literature,
Early" bcsubjectScheme"LCSH" /gt  
ltdcidentifier rdfvalue"isbn0-89887-113-1"
bcidentifierScheme"URN" /gt  
ltbcplacePublicationgtCambridgelt/bcplacePublicatio
ngt   ltdcpublishergtChadwyck-Healeylt/dcpublish
ergt   lt/rdfDescriptiongt   lt/rdfRDFgt
17
Aggregation

• This refers to
• Employing a central knowledge base to gather
  metadata records from different online databases
  using different metadata standards
• Converting heterogeneous metadata records into a
  consistent form
• Developing a range of enhancement services to
  enrich the metadata records gathered.

18
Project using Aggregation - ePrint UK
(Powell 2001)
19
Mapping semantics of different KOSs

• Derivation
• Direct mapping
• Switch language
• Merging
• Co-occurrence mapping
• Satellite and leaf node linking

20
Derivation

• A subject-specific vocabulary is developed based
  on some widely-used general vocabularies.
• Ex MeSH was developed based on the structure of
  LCSH.

21
Direct mapping
(Chan and Zeng 2004)
22
Switch language
(Mai 2003)
23
Projects using a switch language

• The HILT Project
• Uses DDC as a switch language to navigate users
  to find relevant information.
• The Renardus Project.

24
Co-occurrence mapping
(Zeng and Chan 2004)
25
Merging

• Different vocabularies in the same domain can be
  merged into a super-thesaurus.
• Ex The Unified Medical Language System (UMLS)
  merges concepts from about fifty medical
  controlled vocabularies into a metathesaurus.

26
Satellite and leaf node linking

• Editors can select and adapt parts of a general
  vocabulary as a subject-specific vocabulary for
  some particular requirements.
• Ex A number of domain-specific controlled
  vocabularies have been developed by selecting
  parts of LCSH.

27
Ontology mapping for subject cross-search and
browsing

• Current efforts within the digital library
  community include developing ways to map
  different metadata schemes, and ways to map
  different knowledge organisation systems.
• In the semantic web community, the ways to
  improve semantic interoperability include the
  construction of ontology and ontology mapping.
• There is much in common between the methods used
  by these two communities.

28
What is an ontology?

• Definition An ontology is a formal (explicit)
  specification of a conceptualization shared by a
  community of people (R.Studer,1998).
• The difference between an ontology and other
  knowledge organisation systems.

29
Types of ontologies in digital libraries

• Upper level ontology
• Domain ontology.

30
Upper level ontology

• Refers to a common vocabulary including the basic
  concepts, such as things, space, events, time,
  behaviour, etc, and the relations between them
  (Gomez-Perez and Benjamins 1999 Ding and Foo
  2004a).
• Ex OpenCyc, WordNet, and ABC ontology.

31
ABC Ontology

• It provides the notional basis for developing
  domain, role, or community specific ontologies,
  and it incorporates a number of basic entities
  and relationships common across other metadata
  ontologies including time and object
  modification, agency, places, concepts, and
  tangible objects. Communities wishing to build
  their own metadata ontologies and models may then
  extend the ABC entities and relationships as
  needed (Lagoze and Hunter 2001).
• ABC Ontology is designed to incorporate basic
  entities and relationships common across
  different metadata standards, and provide a basis
  to create metadata ontologies, into which
  different metadata schemes can be mapped.

32
OpenCyc Ontology

• This is a universal ontology, in which "every
  concept one can imagine can be correctly linked
  into the OpenCyc Ontology in appropriate places,
  no matter how general or specific, no matter how
  arcane or prosaic, no matter what the context
  (nationality, age, native language, epoch,
  childhood experiences, current goals, etc.) of
  the imaginer" (Stubkjar 2001).
• It provides a framework for further establishing
  custom, and domain-specific ontologies.

33
WordNet Ontology

• This is a manually constructed online lexical
  reference system (Noy and Hafner 1997). In
  WordNet, different lexical objects are organised
  systematically with the basic distinction between
  nouns, verbs, adjectives, and adverbs. Nouns are
  grouped by different concepts, and different
  concepts are organised hierarchically. In
  WordNet, a verb is related to a concepts
  function, and an adjective is related to a
  concepts property.
• The WordNet ontology is often applied to offer a
  taxonomic tree, and also support natural language
  processing.

34
Domain ontology

• A domain-specific vocabulary that encompass the
  concepts in a given domain (such as medical,
  agriculture, computer science, etc) and their
  relationships (Gomez-Perez and Benjamins 1999
  Uschold and Gruninger 1996 Guarino 1997).
• In some cases, potentially, some traditional KOSs
  can be integrated together, and form a basis to
  create a domain ontology.

35
Use of ontologies

• MetaNet
• Different metadata elements from different
  metadata schemes have been mapped to ABC
  ontology.
• Mappings between E-learning object metadata and
  OpenCyc ontology
• Mappings between MeSH and OpenCyc ontology
• Mappings between different subject classification
  systems and OpenCyc

36
An Ontology Library System

• An ontology library system is a library system
  that offers various functions for managing,
  adapting and standardizing groups of different
  ontologies (Ding and Fensel 2001).
• To support searching and browsing different
  ontologies.

37
Conclusion (1)

• A library portal system should be able to
  maximise the reuse of existing library resources,
  such as metadata schemes, and knowledge
  organisation systems.
• In order to improve semantic interoperability, it
  is expected that each resource provider publishes
  metadata schemes, and knowledge organisation
  systems in semantic web enabled format to
  facilitate reusing these resources.
• RDF, XML

38
Conclusion (2)

• In order to facilitate cross-searching
• Develop or apply a common metadata scheme, into
  which different metadata elements from different
  metadata schemes can be mapped.
• Different metadata schemes can also be mapped
  into an upper level ontology.
• These two ways can be developed together.

39
Conclusion (3)

• To facilitate cross browsing by subject
• Different knowledge organisation systems can be
  mapped into a DDC as a subject navigation tree.
• In order to support more powerful computational
  semantics, all concepts, intra-relationships, and
  inter-relationships in different knowledge
  organisation systems can be mapped into an upper
  level ontology.

40
Conclusion (4)

• A variety of mappings have been developed.
• Each type of mapping is designed to offer
  specific capabilities to improve semantic
  interoperability, and limited search or browsing
  functions.
• A combination of the different types of mapping
  is required

41
Thank you and questions!
Libo Si l.si_at_lboro.ac.uk