Accessing Cultural Heritage Collections using Semantic Web Techniques

1
Accessing Cultural Heritage Collections using
Semantic Web Techniques

• Antoine ISAAC
• STITCH Project
• SIKS Semantic Web Seminar, Utrecht
• April 11th, 2007

2
Background

• CATCH_at_ NWO
• Continuous Access To Cultural Heritage
• 10 computer science projects applied to the CH
  field
• Personalization of access, image/text/audio
  analysis
• Integration of projects in CH institutes
  (museums, archives)
• STITCH
• SemanTic Interoperability To access Cultural
  Heritage
• Exchanging and integrating metadata
• Vrije Universiteit, Koninklijke Bibliotheek Max
  Planck Institute

3
Agenda

• Cultural Heritage and Semantic Web
• Two important issues
• Representing Cultural Heritage vocabularies on
  the Semantic Web
• Vocabulary alignment
• Demo

4
Some Needs for CH Collections

• Representation of objects and knowledge about
  them
• Pointing at collection artifacts books
• Describing them creating metadata
• Specific metadata structures (metadata schemes)
• Controlled expert vocabularies (e.g. thesauri)
• Accessing artifacts using metadata
• E.g. search using information contained in
  thesauri

5
KB Illustrated Manuscripts Iconclass vocabulary
6
KB Illustrated Manuscripts
7
Some Needs for CH Collections (2)

• Communicating data to the outside world
• Web portals
• Integrating different collections
• Virtual collections
• The European Library, http//www.theeuropeanlibrar
  y.org
• Geheugen van Nederland, http//www.geheugenvannede
  rland.nl

8
(Biased) Semantic Web

• Pointing at resources documents, knowledge
  objects
• Enabling structured assertions
• Metadata about entities present on the Web
• Using vocabularies with defined semantics
• Ontologies formal definitions of shared
  conceptual vocabularies
• RDF Schema /OWL

ltowlClass rdfabout"Bird"gt
ltowldisjointWithgt ltowlClass
rdfabout"Mammals"/gt lt/owldisjointWithgt
ltrdfssubClassOfgt ltowlClass
rdfID"Animals"/gt lt/rdfssubClassOfgt
lt/owlClassgt ltBird rdfabout"tweety"/gt
9
(Biased) Semantic Web

• Web-based resources allow division/sharing of
• document
• vocabulary
• metadata

http//www.geo.org/voc/

(doc3, hasSubject, Amsterdam)
http//www.kb.nl/eDepot
http//www.ned.nl/doc3
different owners locations
10
Cultural Heritage Collections and Semantic Web

• Categorizing/classifying things
• Structuring descriptions
• Web-based approach
• Semantic Web techniques are good candidates for
  representing and exploiting Cultural Heritage
  metadata

11
Important line of research

• Long-term projects
• MuseumFinland, http//www.museosuomi.fi/
• eCulture, http//e-culture.multimedian.nl/
• Common portals to (many) collections
• Exploiting the data found in the original systems
• Metadata content place, date, creator
• Semantics of vocabularies used to create this
  information
• E.g. hierarchical information
• A Picture featuring a crow features a bird

12
(No Transcript)
13
Agenda

• Cultural Heritage and Semantic Web
• Two important issues
• Representing Cultural Heritage vocabularies on
  the Semantic Web
• Vocabulary alignment
• Demo

14
Representing CH vocabularies on the Semantic Web
- Similarities

• Both ontologies and thesauri bring concept
  hierarchies
• giving the intended meaning of a vocabulary
  through links between its items
• concept/term ? owlClass
• broader ? rdfssubClassOf
• scope notes ? rdfscomment

15
Representing CH vocabularies on the Semantic Web
- Problems

• Thesauri designed for humans, no formal
  interpretation
• How to interpret a thesaurus in RDFS/OWL
• If (Story of) Hercules is a class, what are its
  instances?
• Is Hercules shooting Nessus a subclass of
  Love-affairs of Hercules?
• Thesaurus hierarchy subsumption, mereological
  relation,

16
Representing CH vocabularies on the Semantic Web
Different approaches

• Ontologising
• Cleaning thesaurus by distinguishing roles,
  kinds, etc.
• Cleaning the hierarchical links
• Representing knowledge found in sources as such
• Informal knowledge represented in RDF/OWL formal
  framework

17
SKOS

• Simple Knowledge Organization Systems
• (Future) W3C standard
• Model to represent controlled and structured
  vocabularies on the Semantic Web
• Compatible with community needs
• Core model for representing thesauri,
  classification schemes, etc.

18
SKOS

• Building blocks (ontology) to create XML/RDF data
  about controlled vocabularies
• Classes Concept and ConceptScheme
• Lexical properties
• prefLabel
• altLabel
• Semantic properties
• broader, narrower
• related
• Properties for notes and comments
• scopeNote
• definition

19
SKOS Brinkman Trefwoorden (KB)

• 075607204 geneeskunde
• RT geneesmiddelen
• NT kindergeneeskunde
• 075607220 geneesmiddelen
• UF medicijnen
• 075611791 kindergeneeskunde
• BT geneeskunde
• noot kinderen ouder dan 12 vallen niet onder
• kindergeneeskunde
• medicijnen
• USE geneesmiddelen

20
SKOS Brinkman Trefwoorden (KB)
skos http//www.w3.org/2004/02/skos/corebk
http//www.kb.nl/brinkman/
21
SKOS Brinkman Trefwoorden (KB)
ltskosConcept rdfabout"http//www.kb.nl/brinkman
/bk075607204"gt ltskosprefLabelgtgeneeskundelt/skos
prefLabelgt ltskosrelated rdfresource"http//www
.kb.nl/brinkman/bk075607220"/gt lt/skosConceptgt ltsk
osConcept rdfabout"http//www.kb.nl/brinkman/bk
075607220"gt ltrdftype rdfresource"skosConcept
"/gt ltskosprefLabelgtgeneesmiddelenlt/skosprefLabe
lgt ltskosaltLabelgtmedicijnenlt/skosaltLabelgt lt/sk
osConceptgt ltskosConcept rdfabout"http//www.kb
.nl/brinkman/bk075611791"gt ltrdftype
rdfresource"skosConcept"/gt ltskosprefLabelgtki
ndergeneeskundelt/skosprefLabelgt ltskosbroader
rdfresource"http//www.kb.nl/brinkman/bk07560720
4"/gt ltskosscopeNotegtkinderen ouder dan 12
vallen niet onder kindergeneeskundelt/skosscopeNo
tegt lt/skosConceptgt
22
Agenda

• Cultural Heritage and Semantic Web
• Two important issues
• Representing Cultural Heritage vocabularies on
  the Semantic Web
• Vocabulary alignment
• Demo

23
Cultural Heritage Interoperability Problems

• Problem integrating different databases/metadata
  schemes/vocabularies
• Syntactic interoperability can be solved
• Common format XML (RDF)
• Common vocabulary model (SKOS)
• How about conceptual heterogeneity?

24
The semantic interoperability problem

• There is no standard thesaurus
• We dont really want it
• different vocabularies for different expertise
  domains, traditions, tasks
• Consequence
• klassieke ruïnes vs. landschap met ruïnes
• maagd Maria vs. Heilige Moeder
• Practical problem
• Searching for Heilige Moeder misses maagd
  Maria
• Unless we know both vocabularies

25
Old situation
26
Vocabulary alignment

• STITCH aim find correspondences between
  vocabulary elements
• klassieke ruïnes landschap met ruïnes
• maagd Maria Heilige Moeder

27
New situation
28
Automatic alignment techniques

• Lexical
• Labels of entities and textual definitions
• Structural
• Structure of the formal definitions of entities,
  position in the hierarchy
• Statistical
• Object information (e.g. book indexing)
• Background knowledge
• Using a shared conceptual reference to find links

29
Lexical alignment

• Use preferred labels, synonyms, notes
• Heuristic methods to discover equivalence and
  specialization relations

30
Automatic Alignment Techniques

• Lexical
• Labels of entities and textual definitions
• Structural
• Structure of the formal definitions of entities,
  position in the hierarchy
• Statistical
• Object information (e.g. book indexing)
• Shared background knowledge
• Using a conceptual reference to deduce
  correspondences

31
Statistical alignment
32
Statistic approach Koninklijke Bibliotheek case

• Situation 2 overlapping collections indexed with
  different thesauri
• Comparison means measuring overlap between
  concepts from the thesauri
• Using the sets of books indexed by these concepts
• Results
• 1 9132.9 Schilderijen - schilderkunst
• 2 8088.5 Kwaliteitszorg - kwaliteitsmanagement
• 3 6232.7 Personeelsmanagement - personeelsbeleid
  
• ...
• 17 3421.8 Diabetes mellitus - suikerziekte

33
Agenda

• Cultural Heritage and Semantic Web
• Two important issues
• Representing Cultural Heritage vocabularies on
  the Semantic Web
• Vocabulary alignment
• Demo

34
Demo

• KB Illuminated Manuscripts
• French National Library Mandragore Manuscripts

35
Manuscripts, 2nd Collection BNF Mandragore
36
Manuscripts, 2nd Collection BNF Mandragore
37
Demo

• http//stitch.cs.vu.nl/rp33333/MANDRA-SV-ICE-mandr
  aNewNONE , amphibians
• http//stitch.cs.vu.nl/rp33333/MANDRA-SV-MANDRA-ma
  ndraNewNONE, wheat

38
Conclusion Semantic Web can help Cultural
Heritage

• Representation of collections and associated
  expert vocabularies
• Semantic integration through correspondences
  between different vocabularies
• New opportunities for exploiting cultural
  heritage information

39
Thanks!
40
Links

• Semantic Web at Vrije Universiteit
• http//www.cs.vu.nl/ai/kr/
• http//www.cs.vu.nl/bi/
• SKOS
• http//www.w3.org/2004/02/skos/
• Other Cultural Heritage and Semantic Web projects
• MuseumFinland, http//www.museosuomi.fi/
• eCulture, http//e-culture.multimedian.nl/