Using SKOS in practice, with examples from the classification domain Antoine Isaac Aida Slavic UDC Seminar 2009.

Using SKOS in practice, with examples from the classification domain

Antoine Isaac

Aida Slavic

UDC Seminar 2009

SKOS

• Simple Knowledge Organization System

• SKOS is a model to represent KOSs on in RDF in a simple way– Not aimed at fitting everything!– Not aimed at replacing existing (non-web) formats!

• Now an official W3C recommendation

SKOS basis

SKOS offers a vocabulary for

• Concepts and concept schemes

• Labels

• Semantic relations

• Notes

Networking controlled vocabularies in SKOS

Matching properties as conceptual relations

skos:exactMatch

skos:broadMatch

ex1:cats

ex1:animals ex2:animal

KOS1animalscatswildcats

KOS2animalhumanobject

The relevance of networked KOSs

Johan Stapel, Koninklijke Bibliotheek, Den Haag

Pointers to SKOS deployments

• SKOS implementation reportwww.w3.org/2006/07/SWD/SKOS/reference/20090315/implementation.html

• Library of Congress Subject Headingslcsh.info

• RAMEAU subject headingsstitch.cs.vu.nl/rameau

• DDCdewey.info

SKOS conversion example: NBC

Dutch Basic Classification

21.50 Sculpture: general21.51 Techniques and materials

Representing classes as concepts

rdf:type

nbc:21.51

skos:Concept

21.51 Techniques and materials

Relationships between classes

21.50 Sculpture: general21.51 Techniques and materials

skos:broader

rdf:type

nbc:21.50

skos:Concept

rdf:type

nbc:21.51

skos:Concept

Issues with SKOS representation of classifications

• Captions and labels• Languages• Time versions of classes• Coordination of classes• Auxiliary elements• …

Issue 1: notations, captions and labels

• SKOS enables representing notations• And human-readable preferred (or alternative)

labels, suited for captions

21.51 Techniques and materials

"21.51"^^nbc:notationskos:notation

nbc:21.51

skos:prefLabel"techniques andmaterials"@en

Different choices for notations?

• Can notations be SKOS preferred labels?– They make sense for users & can be displayed

• We can represent them as language-neutral labels

skos:prefLabel "21.51"

"21.51"^^nbc:notationskos:notation

nbc:21.51

skos:prefLabel"techniques andmaterials"@en

Several choices!

Issue 2: Multiple languages

SKOS concepts can have labels in various languages

"21.51"^^nbc:notationskos:notation

nbc:21.51

skos:prefLabel"techniques andmaterials"@enskos:prefLabel

"technieken enmaterialen"@nl

Issue 2: Multiple languages

On dewey.info: one concept per class & language

rdf:type

rdf:type

"641"^^ddc:notationskos:notation

ddc:641/2009/08/about.en

skos:prefLabel"Food & drink"@en

skos:Concept

"641"^^ddc:notationskos:notation

ddc:641/2009/08/about.fr

skos:prefLabel "Nourriture etboisson"@fr

skos:Concept

Several choices!

Issue 2: Multiple languages

Introducing an abstract entity that denotes the time/language-independent class

"641"^^ddc:notationskos:notation

ddc:641/2009/08/about.en

skos:prefLabel "Food & drink"@en

"641"^^ddc:notationskos:notation

ddc:641/2009/08/about.fr

skos:prefLabel "Nourriture etboisson"@fr

ddc:641/

dct:isVersionOf

dct:isVersionOf

Using Dublin Core's isVersionOf link

In RDF, it is possible to use other models in combination with SKOS…

Raising difficult issues: what is a "concept"?

• An artifact– used in descriptions, e.g., as subjects– used as a cluster for different labels with a similar

meaning – in semantic relationships with other concepts

• basis for Knowlegde Organization Systems

• What qualifies as a concept for DDC? For UDC?• Can different choices be done?

Issue 3: Tracking modifications2 RELIGION. FAITHS

21/28 CHRISTIANITY

21 Natural theology. Theodicy. De Deo 22 The Bible. Holy scripture 23 Dogmatic theology24 Practical theology 25 Pastoral theology 26 Christian church in general 27 General history of the Christian church

28 Christian churches, sects 29 NON CHRISTIAN RELIGIONS

NOW.....

2 RELIGION. FAITHS21 Prehistoric and primitive religions22 Religions of the Far East 23 Religions of the Indian subcontinent 24 Buddhism 25 Religions of antiquity 26 Judaism 27 Christianity 28 Islam 29 Modern spiritual movements

Creating representations of the versions

Each time-specific version should be a skos:Concept because:

• They are related to other (versions of) classes • Classification act refers to a time-specific version

A book on the Bible would be wrongly classified by "22" if "22" is the notation for "Far Eastern Religions"

"22"^^udc:notationskos:notation

udcmrf:22_17054

skos:prefLabel"Bible"@en

"22"^^udc:notationskos:notation

udcmrf:22_16544

skos:prefLabel"Far East

religions"@en

dct:isReplacedBy

IF WE FOCUS ON NOTATION... 22

Representing versions of classes

Using Dublin Core's isReplacedBy link

udcmrf:reference/22

"22"^^udc:notationskos:notation

udcmrf:22_17054

skos:prefLabel"Bible"@en

"22"^^udc:notationskos:notation

udcmrf:22_16544

skos:prefLabel "Far Eastreligions"@en

dct:isReplacedBy

dct:isVersionOf

dct:isVersionOf

Notations as a reference

Notations as aggregations

Notations can be seen as hubs for time-specific version of classes

udcmrf:reference/22

"22"^^udc:notationskos:notation

udcmrf:22_17054

skos:prefLabel"Bible"@en

"22"^^udc:notationskos:notation

udcmrf:22_16544

skos:prefLabel "Far Eastreligions"@en

dct:isReplacedBy

ore:aggregates

ore:aggregates

Using the aggregates link

from Object Reuse & Exchange (ORE) initiative

Tracking "Far East religions" in UDC database

CLASS ID: 16544

NOTATION: 22

DESCRIPTION:Religions originating in the Far East

*REPLACES: 299.5 DESCRIPTION: Religion of Oriental peoples ID: 15999

NOTATION HISTORY [22]: YES OLD DESCRIPTION: Bible

There is interesting history information (incl. IDs) in the vocabulary management system!

Concepts across notations

• "Far East religions" existed before: 299.5 • It does not belong to 22 alone

• There is a more abstract Far East Religion entity– its conceptual content may be updated while it keeps

its position in the hierarchy– it might be moved in the notation hierarchy and retain

its conceptual content

"22"^^udc:notation

skos:notation

udcmrf:22_16544

skos:prefLabel

"Far Eastreligions"@en

"299.1"^^udc:notation

skos:notation

udcmrf:299.1_15999

skos:prefLabel

"Religions ofOriental

Peoples"@en

dct:isReplacedBy

Representing concepts across notations

Towards stable abstract concepts?

"22"^^udc:notation

skos:notation

udcmrf:22_16544

skos:prefLabel

"Far Eastreligions"@en

"299.1"^^udc:notation

skos:notation

udcmrf:299.1_15999

skos:prefLabel

"Religions ofOriental

Peoples"@en

dct:isReplacedBy

udc:concept-FarEastReligion

dct:isVersionOf dct:isVersionOf

udcmrf:reference/22

"22"^^udc:notationskos:notation

udcmrf:22_17054

skos:prefLabel"Bible"@en

"22"^^udc:notation

skos:notation

udcmrf:22_16544

skos:prefLabel

"Far Eastreligions"@en

dct:isReplacedBy

ore:aggregates

ore:aggregates

"299.1"^^udc:notation

skos:notation

udcmrf:299.1_15999

skos:prefLabel

"Religions ofOriental

Peoples"@en

dct:isReplacedBy

udc:concept-FarEastReligion

dct:isVersionOf dct:isVersionOf

Two different versioning & aggregation mechanisms at work

Conclusion

• SKOS is an interoperability enabler• There are still issues

Different conversions reveal different choices

• But that's not a new problem, reallyStandardization often needs best practices & use cases

• Now, it is easier to publish and compare dataAs data is (will be) linked, it is even crucial to do it!

Interaction between information experts and techies is crucial

Thank you!

• Questions?

Drawing inspiration from• Joseph Tennis & Stuart Sutton• Michael Panzer• Jakob Voss