How to De-engineer a Semantic Web:

How to De-engineer a How to De-engineer a Semantic Web:Semantic Web:Some thoughts on Linking Archaeological Some thoughts on Linking Archaeological DataData

Leif Isaksen, Graeme Earl & Kirk MartinezLeif Isaksen, Graeme Earl & Kirk MartinezUniversity of SouthamptonUniversity of Southampton

The ProblemThe Problem

Summary dataSummary data

Fragments of MeaningFragments of Meaning

Instance dataInstance data

Mixing ModelsMixing Models

ContextContext FindFind Quantity Quantity (weight)(weight)

1717 Dressel 2-4 Dressel 2-4 CatalanCatalan 400400

1818 Dressel 2-4 Dressel 2-4 EgyptianEgyptian 150150

cont.cont. findsfinds no.no. dimdim..

reg.?reg.?

Floor Floor AA

Egyptian Egyptian amph.amph. 33 rimsrims yesyes

Floor Floor BB

Cat.Cat. 11 vol.vol. nono

TypeType

RelationshipRelationship

Mixing ModelsMixing Models

-- -- --

abab blah-de-blahblah-de-blah xyzxyz

cdcd da-de-dada-de-da uvwuvw

-- -- -- -- --

dfgdfsdfgdfs Tum-Te-TumTum-Te-Tum ii bishbish 11

ukuikukuik Tiddly-PomTiddly-Pom jjbosboshh

00

TypeType

RelationshipRelationship

The representation The representation is not the data!is not the data!

ApproachesApproaches

TRANSLATIONTRANSLATION(TRansparent Negotiation and Sharing of Local Application Terminologies, (TRansparent Negotiation and Sharing of Local Application Terminologies, Instances and ONtologies)Instances and ONtologies)

Peer-to-peerPeer-to-peerServer frameworkServer framework

Problems with the Problems with the TRANSLATION approachTRANSLATION approach

ClosedClosed

An engineers-eye viewAn engineers-eye view

‘‘Internet thinking’ not ‘web thinking’Internet thinking’ not ‘web thinking’

The network effect is created by Web 2.0 not The network effect is created by Web 2.0 not Internet 2.0!Internet 2.0!

Linked DataLinked Data

http://linkeddata.org

GeoNamesGeoNames

SubgraphsSubgraphs

Canonical TypesCanonical Types

Instance DataInstance Data

Extension/IntensionExtension/Intension

What is ‘Dressel 20’?What is ‘Dressel 20’?

Defined by attributes Defined by attributes (Intension -> a priori (Intension -> a priori analysis)?analysis)?

Defined by examples Defined by examples (Extension -> a (Extension -> a posteriori analysis)?posteriori analysis)?

Quantity vs. uncertaintyQuantity vs. uncertainty

Quantity vs. uncertaintyQuantity vs. uncertainty

Quantity vs. uncertaintyQuantity vs. uncertainty

Interface - ease of useInterface - ease of use

Archaeological data is distributed in fragments Archaeological data is distributed in fragments amongst (generally) non-tech literate amongst (generally) non-tech literate professionals with no money and less time!professionals with no money and less time!

Conversion must beConversion must be

SimpleSimple

FastFast

BeneficialBeneficial

ImplementationImplementation

De-engineering the dataDe-engineering the data

4 Steps4 Steps

Define OntologyDefine Ontology

Define Canonical TypesDefine Canonical Types

Map local termsMap local terms

Generate RDFGenerate RDF

Two OntologiesTwo Ontologies

Linked DatasetsLinked Datasets

archvocab.netarchvocab.netarchvocab.netarchvocab.net

archaeology.rkbexplorer.archaeology.rkbexplorer.comcom

archaeology.rkbexplorer.archaeology.rkbexplorer.comcom

Port Networks Triplestore +Port Networks Triplestore +Port Networks partner sitesPort Networks partner sitesPort Networks Triplestore +Port Networks Triplestore +Port Networks partner sitesPort Networks partner sites

SKOS SKOS SchemesSchemes

SKOS SKOS SchemesSchemes

Additional Additional RDFRDF

InstanceInstancedatadata

archvocab.net/amphoraarchvocab.net/amphora

archaeology.rkbexplorer.archaeology.rkbexplorer.comcom

Project-specific data at class levelProject-specific data at class level

class equivalence using SKOS class equivalence using SKOS (skos:exactMatch)(skos:exactMatch)

class origin (e.g. Dressel 20 :hasOrigin class origin (e.g. Dressel 20 :hasOrigin Baetica)Baetica)

other infoother info

Helps separate vocabulary (archvocab.net) Helps separate vocabulary (archvocab.net) from politics!from politics!

DemoDemo