21
D15.3 Report on Semantic Annotation and Linking Version 1.0 January 2017 Grant Agreement number: 313193 Project acronym: ARIADNE Project title: Advanced Research Infrastructure for Archaeological Dataset Networking in Europe Funding Scheme: FP7-INFRASTRUCTURES-2012-1 Project co-ordinator: Franco Niccolucci, PIN Scrl - Polo Universitario "Città di Prato" Tel: +39 0574 602578 E-mail: [email protected] Project website address: http://www.ariadne-infrastructure.eu ARIADNE is a project funded by the European Commission under the Community’s Seventh Framework Programme, contract no. FP7-INFRASTRUCTURES-2012-1-313193. The views and opinions expressed in this report are the sole responsibility of the authors and do not necessarily reflect the views of the European Commission.

D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

  • Upload
    others

  • View
    9

  • Download
    0

Embed Size (px)

Citation preview

Page 1: D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

D15.3ReportonSemanticAnnotationandLinking

Version1.0

January2017

GrantAgreementnumber: 313193

Projectacronym: ARIADNE

Projecttitle: AdvancedResearchInfrastructureforArchaeologicalDatasetNetworkinginEurope

FundingScheme: FP7-INFRASTRUCTURES-2012-1

Projectco-ordinator: FrancoNiccolucci,PINScrl-PoloUniversitario"CittàdiPrato"

Tel: +390574602578

E-mail: [email protected]

Projectwebsiteaddress: http://www.ariadne-infrastructure.eu

ARIADNEisaprojectfundedbytheEuropeanCommissionundertheCommunity’sSeventhFramework Programme, contract no. FP7-INFRASTRUCTURES-2012-1-313193. The viewsandopinionsexpressedinthisreportarethesoleresponsibilityoftheauthorsanddonotnecessarilyreflecttheviewsoftheEuropeanCommission.

Page 2: D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

ARIADNE–D15.3:ReportonSemanticAnnotationandLinking PreparedbyUSW

ARIADNE 2 January2017

Aboutthisdocument

This document is a contractual deliverable of the ARIADNE project. The research leading to theseresultshasreceivedfundingfromtheEuropeanCommunity'sSeventhFrameworkProgramme(FP7-INFRASTRUCTURES-2012-1) under grant agreement n° 313193. But the content of this documentcannotbeconsideredtoreflecttheviewsoftheEuropeanCommission.

Partnerinchargeofthedeliverable:

UniversityofSouthWales(USW)

Authors: DouglasTudhope(USW)CeriBinding(USW)

Contributors: DimitrisGavrilis(ATHENARC)PaulBoon,HellaHollander(KNAW-DANS)JeremyAzzopardi(SND)AndreasVlachidis(USW)GuntramGeser(SRFG)

Qualityreview: FrancoNiccolucciandPaolaRonzino(PIN)

Page 3: D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

ARIADNE–D15.3:ReportonSemanticAnnotationandLinking PreparedbyUSW

ARIADNE 3 January2017

Tableofcontent

1 ExecutiveSummary....................................................................................................................42 SemanticannotationandlinkingwithinARIADNE......................................................................5

2.1 Semanticlinkinginthespatialdimension..................................................................................52.2 Semanticlinkinginthetemporaldimension..............................................................................52.3 Semanticlinkinginthesubjectdimension.................................................................................62.4 Semanticlinkingtoolsinthesubjectdimension........................................................................7

2.4.1 VocabularymappingtoolsrelevanttoWP15................................................................................72.4.2 VocabularymappingtoolsinARIADNE..........................................................................................92.4.3 VocabularymappingsconductedforARIADNE............................................................................112.4.4 VocabularymappingswithintheARIADNEportal.......................................................................12

2.5 Wood/Dendrochronologycasestudyonitemleveldataintegration......................................133 Summaryandlessonslearned..................................................................................................174 References...............................................................................................................................19

Page 4: D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

ARIADNE–D15.3:ReportonSemanticAnnotationandLinking PreparedbyUSW

ARIADNE 4 January2017

1 ExecutiveSummary

Briefsummary

Thereviewanddevelopmentworkdescribedinthisreportfocusesontheaspectsofsemanticlinkingand annotation particularly relevant to ARIADNE. Semantic linking within ARIADNE is consideredwithinthespatial, temporalandsubjectdimensions.Thesubjectdimension isconsidered indepth,startingwithareviewoflinkingtoolsconsideredrelevanttoARIADNEfollowedbyadiscussionoftheARIADNEapproachandthevocabularymappingtoolsusedwithinARIADNE.TheGettyAATprovedanappropriatevocabularymappinghubthataffordedamultilingualsearchcapabilityintheARIADNEPortalviathesemanticenrichmentofpartnersubjectmetadatawithderivedAATconcepts.

A case study conducted an exploratory investigation of the semantic integration of extracts fromarchaeological datasets with information extracted via NLP across different languages. Theinvestigationfollowedabroadthemerelatingtowoodenmaterialincludingshipwrecks,withafocusontypesofwoodenmaterial,samplestaken,woodenobjectswithdatingfromdendrochronologicalanalysis,etc.TheDemonstratorisavailableforgeneraluse.Theuserisshieldedfromthecomplexityof the underlying semantic framework (based on the CIDOC CRM and Getty AAT) by the Webapplicationuserinterface.TheDemonstratorhighlightsthepotentialforarchaeologicalresearchthatcaninterrogategreyliteraturereportsinconjunctionwithdatasets.Queriesconcernwoodenobjects(e.g. samplesofbeechwoodkeels), optionally fromagivendate range,withautomaticexpansionoverhierarchiesofwoodtypes.

Lessonslearned

o The spatial, temporal and subjectdimensions are key to archaeology and themain vehicle forsemantic linking. Cleansing and normalisation via semantic enrichment is necessary in eachdimension.

o The Getty AAT proved an appropriate vocabulary mapping hub that afforded a multilingualsearchcapabilityintheARIADNEPortal.

o TheWood/Dendrochronologycasestudydemonstratedthefeasibilityofconnectinginformation(at a detailed level) extracted fromdatasets and grey literature reports in different languagesand semantic cross-searching of the integrated information. The case study suggests that thesemantic linking of textual reports and datasets opens up possibilities for integrativearchaeologicalresearchacrossdiverseresources.

o The Wood/Dendrochronology demonstrator shows that a Web application can hide thecomplexity of the underlying semantic framework from the user and allow querying andbrowsing the information without expertise in SPARQL; it illustrates that more intuitive userinterfacesarepossibleforsearchingRDFdatasetsthantheusualSPARQLendpointorbrowsinghyperlinks.

Page 5: D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

ARIADNE–D15.3:ReportonSemanticAnnotationandLinking PreparedbyUSW

ARIADNE 5 January2017

2 SemanticannotationandlinkingwithinARIADNE

Semantic annotation is covered in part by the ARIADNE Natural Language Processing (NLP) workreportedunderWP16.Formoreinformationonautomatictextannotation(andtext-mining)withinarchaeology,includingmachinelearningoutcomesfromtheADSArchaeoToolsProjectandtherule-based OPTIMA toolkit from the USW STAR Project, see Richards et al. (2015) and the deliverableARIADNED16.4-FinalReportonNaturalLanguageProcessing,whichalsoreportsonNLPoutcomesfrom ARIADNE. One strand of these outcomes contributed to the Wood/Dendrochronology dataintegrationcasestudyandDemonstrator,whichcombinessemanticlinksbetweendatasetsand(viaNLP)archaeologicalreports.Thisisdiscussedbelow.

In archaeology, semantic linking generally involves connections (usually semi-automatic withintellectual review) between archaeological resources via an intermediary knowledge organizationsystemcoveringrelevantaspectsofthespace,timeandsubjectdimensions.Semanticlinkinginthesubjectdimension isdiscussedbelowas regardsmappingbetweensubject thesauriandcontrolledvocabularies,while linkingvia theCIDOCCRMcoreontology, isdiscussed inARIADNED14.2 -PilotDeploymentExperiments.

2.1 Semanticlinkinginthespatialdimension

ForARIADNE, a key componentof the strategy for the spatial domainhasbeen theadoptionof astandardformatforspatialcoordinates,WGS84.DifferentpartnersemploydifferentlocalcoordinatesystemswhicharenormalisedtoWGS84.NormalisedspatialcoordinatesconstitutethemainspatialaccessmethodintheARIADNEPortal.Relevantgazetteersofplacenames,includingGeoNamesandTGN, are discussed in ARIADNED3.1 - Initial report on standards and on the project registry. Thework of the Pelagios project has been a significant contributor to theDigital Humanities arena ofspatiallinkeddata,includingvariousworkshopsintheLinkedPastsseries.PelagiosmakesuseofthePleiades gazetteer1 (and its URIs) to connect online resources that refer to places in the ancientworld via LinkedOpenData. Pelagios does not attempt to define a complex datamodel, rather itseeks to offer a uniformway to build links betweendifferent gazetteers via theOpenAnnotationOntology.Pelagiosmetadataaimstosupportinteroperabilitywhileimposingminimaloverheadsondataproviders.Pelagiosisdevelopingvarioustools(betaattimeofwriting),includingRecogito2,anannotationplatformforannotatingimages,creatingmaps,linkingresearchdatatothewiderwebofdata,thePelagiosMapTilesmappingapplication,thePeripleosearchserviceandAPI3.VariousothereffortsareongoinginthesecondphaseofthePelagiosproject,whichaimstoencouragecommunityparticipation in the enterprise: “Pelagios Commons provides online resources and a communityforumforusingopendatamethodstolinkandexplorehistoricalplaces”4.

2.2 Semanticlinkinginthetemporaldimension

In the temporal domain, ARIADNE made the decision to adopt the recently emerged PeriodOgazetteerasa centralhub forexpressing standardperioddefinitions, inorder to linkandvisualize1Pleiades,https://pleiades.stoa.org2Recogito,http://recogito.pelagios.org3Peripleo,https://github.com/pelagios/peripleo#peripleo-api4Pelagios,http://commons.pelagios.org

Page 6: D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

ARIADNE–D15.3:ReportonSemanticAnnotationandLinking PreparedbyUSW

ARIADNE 6 January2017

time period data. “PeriodO is a gazetteer of scholarly definitions of historical, art-historical, andarchaeologicalperiods.Iteasesthetaskoflinkingamongdatasetsthatdefineperiodsdifferently.Italso helps scholars and students see where period definitions overlap or diverge” (PeriodO)5. A‘period’ is of course a complex archaeological concept, which can encompass timespans, spatialextents and sometimes cultural forms (Niccolucci&Hermon 2015). PeriodOdefines a datamodelthat includes a name for the period, some temporal bounds, an association with a geographicalregion and which has some literary warrant (Shaw 2015; Shaw et al. 2015). The PeriodO Client6allowsuserstobrowsethePeriodOdatasetofperiods,comparedifferentperioddefinitions,extracttheURIforaperiod,extractsemanticallystructureddata(JSON-LD),submitnewperiodsorupdateexistingonestheuserhaspreviouslyentered. ARIADNEpartnersexpressedtemporalmetadataforarchaeologicalperiodsusinglocalvocabularies,withstartandenddatesforeachterm.Theunifiedlist of ARIADNE period vocabularies was made available to PeriodO in a collaboration with thatproject.ThisresultedintheARIADNEcollectionofperioddefinitionsinPeriodO7,whereURIsidentifyeachperiodanddistinguishthemeaningofaperiodnameindifferentplaces.

2.3 Semanticlinkinginthesubjectdimension

MuchoftheeffortofWP15hasconcernedsemanticlinkinginthesubjectdomain.Mappingbetweensubjectthesauriandothercontrolledvocabulariesisanimportantcontributortointeroperabilityandcross search indisciplineswithdiversedatasets, suchasarchaeology,and isparticularly importantformultilingualcapability(seethereviewinZeng&Chan2004).ARIADNED15.1givesanoverviewofvocabulary mapping drawing on the ISO thesaurus standard (ISO 25964-2:2013), discussing therationaleof vocabularymappingwithinARIADNEand thechoiceof theGettyArtandArchitectureThesaurus (AAT)8 as amapping hub. TheAAT is available as LinkedOpenData in SKOS, publishedunder the Open Data Commons Attribution License (ODC-By) 1.09. The AAT contains over 40,000concepts and over 350,000 terms, organised in seven facets (and 33 hierarchies as subdivisions):Associatedconcepts,Physicalattributes,Stylesandperiods,Agents,Activities,Materials,Objectsandoptional facets for time and place (Harpring 2016). The AAT’s scope is broader than archaeology,encompassing visual art, architecture, othermaterial heritage, archaeology, conservation, archivalmaterials,etc.,butcontainsmanyusefulhighlevelarchaeologicalconcepts,particularlyintheBuiltEnvironment,MaterialsandObjectshierarchies.

DifferentpartnersubjectvocabulariesaremappedtotheAAT,whichthenformsthebasisforsearchacross indexing vocabularies in different languages. Themajor exercisemapping between partnernativevocabulariesandtheAATasacentralhubisdescribedinD15.1, includingreflectionsonthemappingprocess. The creationof linksdirectlybetween the items fromdifferent vocabularies canbecomeunmanageableasthenumberofvocabularies increases.Ascalablesolution is toemployahub architecture, an intermediate structure on to which concepts from ARIADNE data providersource vocabularies can bemapped. The benefits of vocabularymapping are clear for collectionswithmultilingualmetadataandcontent.ByusingtheAATasacentralhub,asemanticsearchonanAAT concept can retrieve results originally indexed by terms from various languages (Binding and

5PeriodO,http://perio.do,https://github.com/periodo6PeriodOclient,http://n2t.net/ark:/99152/p07ARIADNEcollectionofperioddefinitionsinPeriodO,http://n2t.net/ark:/99152/p0qhb668GettyArtandArchitectureThesaurus,http://www.getty.edu/research/tools/vocabularies/aat/9GettyVocabulariesasLinkedOpenData,http://www.getty.edu/research/tools/vocabularies/lod/index.html

Page 7: D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

ARIADNE–D15.3:ReportonSemanticAnnotationandLinking PreparedbyUSW

ARIADNE 7 January2017

Tudhope2016).Thiscanpotentially improverecall (indexing inseveral languages)andprecision(inliteralstringsearch,falseresultsmayarisefromhomographs).

2.4 Semanticlinkingtoolsinthesubjectdimension

ARIADNED15.3isconcernedwiththetoolsemployedforsemanticlinking.Whilesomeveryspecificlinks between specific datasets, or specific cross references between reports, will exist, mostsemantic linking in the subjectdimension is achievedvia the coreontology (CIDOCCRM)and theGetty AAT, which have been adopted as ARIADNE linking standards. Before describing the toolsemployed for mapping vocabularies in ARIADNE, we briefly review other relevant vocabularymappingtools.

2.4.1 VocabularymappingtoolsrelevanttoWP15

TheVocBenchpublishingplatformoftheUN’sFAOhasseenamajorSKOSLinkedDataeffortintheagricultural domain,mapping themultilingual AGROVOC thesaurus to (at the last count) 13 otherthesauri (Caracciolo et al. 2012 and 2013), including LCSH (Library of Congress Subject Headings),GEMET(GeneralMultilingualEnvironmentalThesaurus)andSTW(StandardThesaurusforEconomics/StandardThesaurusfürWirtschaft).Thishasbeena longtermproject;earlywork isdiscussedbyLiangandSini(2006).

InARIADNE,itispossibletoassociateAATconceptsdirectlywithpartnersubjectmetadatabytakingaccount of mappings from partner vocabularies to the AAT created for ARIADNE. In addition tosemanticlinkingbymeansofmappingbetweenvocabularies,itissometimesnecessarytoassociateconceptsfromcontrolledvocabularies(viaURIUniformResourceIdentifiers)withsubjectmetadatapresentedonlyasfreetext.Forexample, inthecaseofaproject likeEuropeana,whichsometimeshas to ingest subject metadata without knowledge of the original native vocabulary (if any), it issometimes desirable to associate a controlled vocabulary concept (e.g. from the AAT) withpotentially ambiguous, free text metadata, without being able to rely on a previous vocabularymapping. All that may be available on the source side is the free text subject label without anyfurthercontexttotakeintoaccountfromasourcevocabulary.

Inboththesesituations,theprocessofassociatingacontrolledvocabularyconceptiscalledsemanticenrichment. Enrichment can be considered a special case of semantic annotation. For recentdevelopments in Europeana of semi-automatic 'enrichment' of free text metadata elements withAAT URIs, see Charles et al. (2014). Stiller et al. (2014) report on a Europeana case study ofenrichment with discussion of best practice. A Europeana Data Model case study describes anexample of the enrichment of Europeana data with AAT (Charles and Devarenne 2014). This canbring benefits for multilingual capability, for example by automatically changing the subjectmetadatalabeldisplayedifthelanguageoftheuserinterfaceischangedbytheuser.

A specific MORe enrichment service for deriving AAT concepts from partner vocabularies wasdeveloped as part of ARIADNE utilising the AAT vocabulary mappings produced in WP15 (see'VocabularymappingsconductedforARIADNE'belowandARIADNED15.1-ReportonThesauriandTaxonomies). Various enrichment services have been developed within the MORe aggregationframeworkfortheLoCloudproject(Meghinietal.,forthcoming).Theseinclude:

o GeocodingbasedonGeoNames;

o Semantic enrichment,where a vocabularymatching service suggests SKOS concepts basedontitle,descriptionsandsubject-relatedinformationinametadatarecord;

Page 8: D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

ARIADNE–D15.3:ReportonSemanticAnnotationandLinking PreparedbyUSW

ARIADNE 8 January2017

o WikipediaandDBPediaautomaticenrichment:avariationof thesemanticenrichmentthatsuggestsWikipediaandDBPediaentries,basedonthemetadatarecord;

o Language identification: Identifies languages based on a title or description using ApacheTika;

o Thesaurimappings: Allows loading andmanaging SKOS conceptsmappings from SKOSifiedsubjecttermstoatargetSKOSthesaurus.

The Instance Matching application (SAIM10) is a browser based interface intended to supportinstancematchingonRDFdata, generally though this is somewhatoutside the focusof theWP15work.ConsiderationwasalsogiventotheFP7supportedLinkedDataIntegrationFramework(SILK11).SILK is a sophisticated, general semantic web framework for creating links and not designedspecificallyformappingthesauri.ASILK‘linkspecification’allowsthecomparisonofpreferredlabelsfrom two thesauri (using the Levenshtein distance algorithm). SILK is a complex environment andthere isan inevitable learningcurve inorder tocreate linkage rules, transformations,aggregationsand in the interpretation ofmatch scores.While the SILK application is interactive and allows forsome comparison, it can be argued that themain focus of such tools tends to be automatic linkgeneration functionality and bulk automation. The user is not given convenient contextualinformation(specifictothethesauri)tofacilitateinformedjudgmentsonthecorrectnessofpotentialmappings.

Twoarchaeologicaldomainpresentationsatthe14thEuropeanNetworkedKnowledgeOrganizationSystems (NKOS)Workshop, at TPDL 2015 in Poznań discussedwork on SKOS vocabularymappingusingAMALGAME (AmsterdamAlignmentGenerationMetatool12), developed at FreeUniversity ofAmsterdam within the ClioPatria framework. Stahn (2015) made a pilot study with three DAIvocabularies,expressingthemasSKOSandthenmappingviaAMALGAME.Shemakesthepointthatthe matching is on string level only. Kempf (2015) conducted experiments at ZBW (LeibnizInformationCentreforEconomics)involvingtheGermanSTWThesaurusforEconomicsandanexactlanguagedependentstringmatchonterms.Asecondphaseexploreddifferenttypesofenrichmentofthethesaurusdatabeforethemappingprocess,suchasdescriptors,synonyms,resultsfromothermappings. Theconclusionswerethatstringmatchcouldnottakeaccountofstructuraldifferencesbetweenthesourceandtargetvocabularycontextsbutthattheexperimentenrichingthevocabularyterminologybeforeattemptingmappingofferedusefulpotential candidates for further intellectualmapping.AMALGAMEgivesachoiceofsimilaritymetricsandpropertiestomatchon.Howevertheprocessisthenbasedonautomaticstringmatch(althoughtheuserdoesdecidewhethertoacceptamapping) and does not take account of the hierarchical context. The work with AMALGAME iscontinuingunderthenewprojectCultuurlink13,theDutchCulturalHeritageHub,whereanalignmentservice is offered for SKOS thesauri and various thesauri are available. A complex, interactivealignmentstrategygraphicaleditorisavailable.ThisemploystheAMALGAMEmappingsystemandafilterisavailablethattakessomeaccountofhierarchicalcontext.

10SAIM,http://saim.aksw.org11SILK,http://wifo5-03.informatik.uni-mannheim.de/bizer/silk12AMALGAME,http://semanticweb.cs.vu.nl/amalgame/13Cultuurlink,http://cultuurlink.beeldengeluid.nl/app/#/start

Page 9: D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

ARIADNE–D15.3:ReportonSemanticAnnotationandLinking PreparedbyUSW

ARIADNE 9 January2017

2.4.2 VocabularymappingtoolsinARIADNE

Vocabularymappingisnotatrivialexercise.Highqualitymappingrequiresdomainexperts,whomaynothaveexpertiseincomputingsemantictechnologies.Thevocabulariesthemselvescanvaryfromasmallnumberofkeywordsfromapicklist foraparticulardatasettostandardnationalvocabularieswithalargenumberofconcepts.Ingeneralatthepresenttime,fullyautomaticvocabularymappingdoes not currently delivermappings of sufficient operational quality; the ARIADNE approach is tofocus on quality mappings, in order to support cross search at a medium level of generality.Archaeologydomainvocabularyproviderstendtohavelimitedresourcestospendonthetechnicalaspectsofvocabularymappingtools.Itisimportanttosupportexpertintellectualreviewandprovideenoughcontextualdatatoallowvocabularyproviderstomakeaninformeddecisionontheproposedmappings.Forexample,itmightbeusefulfortheusertocomparethesaurusstructuressidebyside.

In general for ARIADNE purposes, tools should be light weight, user centered and focused onthesaurustothesaurusmapping,inordertofacilitateuser(vocabularyprovider)involvementinthemapping process. In some cases, ARIADNE partners drew on their own resources to create thevocabularymappings toAAT, for example by directly browsing and searching the local vocabularyand the AAT or using local utilities (see ARIADNED15.1). In addition, two lightweight toolsweredevelopedbyUSWtosupportthedomainexpertsdoingthemappingbetweenvocabularyconcepts,orientedtodifferentcontextsforthevocabularies.

An interactive vocabulary mapping tool was developed for ARIADNE that enables archaeologysubject experts create SKOSmapping relationships between local vocabularies available online asSKOS and the AAT (Binding& Tudhope 2016). The lightweightWeb application presents conceptsfrom chosen source and target vocabularies side by side, thus presenting context to allow aninformed choice when deciding on potential mappings. The tool is designed for vocabulariesexpressedinRDF/SKOS;itqueriesexternalSPARQLendpointsratherthanstoringlocalcopiesofthevocabularies.SKOSmappingrelationships(Miles&Bechofer2009)areemployedforexpressingthemappings, (e.g. skos:broadMatchor skos:closeMatch). The resulting set ofmappings canbe savedlocally,reloadedandexportedtoanumberofdifferentoutputformats(JSONforuseinARIADNE).

Figure 1 shows an example where a source concept, castle, has just been mapped to the AATconcept, castles (fortification). The right hand pane shows various AAT concepts that potentiallymatchonthepreferredterm.TheuserhasselectedtheAATconcept,castles(fortification)andboththe broader concept (fortifications) and the narrower concepts (châtelets,moated castles,motte-and-baileycastles,qasrs)aredisplayed,showingthesemanticcontextinaconcisemanner.Eachoftheseadditional concepts can themselvesbebrowsed in the tool toexplore the semantic context.Themappinghasbeenconsideredaclosematchbytheuserandtheresultingmappingisdisplayedatthebottomofthescreen,togetherwithtwootherrecentmappings.ThetoolisavailableasopensourcewiththecodefreelyavailableonGitHub14.

14VocabularyMatchingToolsourcecodeforlocaldownloadandinstallation,

https://github.com/cbinding/VocabularyMatchingTool

Page 10: D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

ARIADNE–D15.3:ReportonSemanticAnnotationandLinking PreparedbyUSW

ARIADNE 10 January2017

Figure1:Vocabularymatchingtool

Asecondmappingapproachwasdevelopedforsourcevocabularies thataresmaller term listsandnotyetexpressedinRDF.Suchtermlistsareoftenavailableoreasilyrepresentedinaspreadsheet.Astandard template with example mappings was designed to support domain experts in theintellectualmapping of terms to the target vocabulary, together with the XSLT conversion of thespreadsheetdatatoJSONandNTriplesformatsforuploadtoARIADNE.ThefinaltemplateisinpartageneralisationofthemappingspreadsheetscreatedbyDANSandSNDfortheirmappingwork.ACSVtransformationproducestherepresentationofthemappingsinRDF/JSONformatandisavailableasopen source15. The user completes themapping templatewith the set of source and target (AAT)concepts. The XSLT transformation expresses the mappings in JSON and NTriples formats. Themappingtemplatecontainedatabtorecordmetadataforthemapping.Infuturework,makingthemappings available as outcomes in their own right, with appropriate metadata for the mappingswouldbedesirable.

Thestandardspreadsheetwasaccompaniedbyasetofguidelinesinformedbyapilotexercise(withadditional support from the vocabulary teamon problematicmappings or existing precedents). Insomecases,data cleansingwas requiredbefore themappingexercise couldproceed. In fact, datacleansingwasnecessaryatvariousstagesofWP15work.Themaintoolused fordatacleansingbyUSWwastheopensourcetoolOpenRefine(formelyGoogleRefine).

15ARIADNEsubjectmappings:Spreadsheettemplateandconversion,https://github.com/cbinding/ARIADNE-

subject-mappings

Page 11: D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

ARIADNE–D15.3:ReportonSemanticAnnotationandLinking PreparedbyUSW

ARIADNE 11 January2017

2.4.3 VocabularymappingsconductedforARIADNE

Figure 2 summarises the mappings produced. In total 6416 mappings were conducted, withmappingsbyindividualpartnersrangingfromafewtoover1600terms.ByDecember2016,conceptsfrom 27 vocabularies employed by 12 project partners have beenmapped to the AAT; 17 of thevocabulary mappings were conducted with the spreadsheet template (or a similar partnerspreadsheet), 2 using the online interactive mapping tool (i.e. when the source vocabulary wasavailableinRDF/SKOS)and8usingthepartner’sown(intellectual/manual)resources.

Figure2:MappingsconductedaspartofWP15

The mappings were converted by USW to JSON format for use in the ARIADNE Registry to helpaugment partner subject metadata with appropriate AAT concepts. The ARIADNE data catalogue

Page 12: D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

ARIADNE–D15.3:ReportonSemanticAnnotationandLinking PreparedbyUSW

ARIADNE 12 January2017

employs theMoRe(Metadata&ObjectRepository)aggregator16 toharvest themetadataprovidedbytheprojectpartnersutilisingtheOpenArchivesInitiativeProtocolforMetadataHarvesting(OAI-PMH). A bespoke AAT subject enrichment service has been developed by DCU that imports thepartner vocabularymappings and applies them to thepartner subjectmetadata, deriving anAATconcept (both preferred label and URI) to augment the existing subject metadata in the datacatalogue.ThecataloguemetadataissuppliedtotheARIADNEportal,wherethesearchfunctionalitycanretrievemultilingualrecordsrelatedviatheAATconcepts.

2.4.4 VocabularymappingswithintheARIADNEportal

TheenrichedAATconceptsaffordamultilingualcapabilityintheARIADNEPortal,allowingasearchonanAATconcepttoreturnresultsfromsubjectmetadataindifferentlanguages.

Figure3:PortalQueryonAATsubject:SettlementsandLandscapesshowingresultsfromAIAC(FastiOnline),INRAPandDANS,withmultiplelanguages(December2016).

Figure3showsaqueryonthePortalmakinguseofthemappings.OnthemainResultsscreen,asetof filters isavailable forrefiningasearchfollowingthefacetedsearchparadigm.Onefilter,namedSubject,ispopulatedbytheMoReenrichmentprocessdescribedabove;effectivelytheSubjectsareAAT concepts,whichhavebeenmapped from thenative vocabulary subjectmetadataof thedata16MoRe(Metadata&ObjectRepository)aggregator,http://more.dcu.gr

Page 13: D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

ARIADNE–D15.3:ReportonSemanticAnnotationandLinking PreparedbyUSW

ARIADNE 13 January2017

resources inthePortal.Figure3showsthatasimplequeryonthesingleAATconcept,Settlementsand Landscapes, is able to retrieve results inmultiple languages for records originating fromAIAC(Fasti),INRAPandDANS.

In addition to semantic linking within the metadata of the ARIADNE Portal, work in WP15 alsoexplored semantic linking at themore detailed level of individual itemswithin ARIADNE datasets.Extracts from the datasets were additionally linked to archaeological reports as a result of NLPsemanticannotationworkperformedaspartofWP16.Thesemanticframeworkforthelinkingwasthe combined CIDOC CRM and Getty AAT. The following section discusses a case study thatinvestigatedthisapproachandtheresultingDemonstrator.

2.5 Wood/Dendrochronology case study on item level dataintegration

AjointWP15andWP16effortconductedanexploratoryinvestigationofthesemanticintegrationofextracts from archaeological datasets with information extracted via NLP (Natural LanguageProcessing) across different languages. The case study17 was based on a loose theme ofarchaeological interest inwoodenobjectsandtheirdatingviadendrochronologicaltechniques.TheworkwasundertakenbyUSWonthetechnicalside,incollaborationwithDANSandSNDasregardsDutchandSwedisharchaeologicaldatasets,reportsandvocabularies.

The case study investigateswhether it is possible to achieve adegreeof semantic interoperabilitybetween archaeological datasets and data derived from applying NLP information extractiontechniques to the textual content of grey literature reports in different languages. The semanticframeworkcombinedtheCIDOCCRMwiththeGettyAAT.Thecasestudyhasabroadthemerelatingtowoodenmaterial includingshipwrecks,witha focuson indicationsof typesofwoodenmaterial,samples taken,wooden objectswith dating fromdendrochronological analysis, etc. The resourcescomprisefiveEnglishandDutchlanguagedatasetsandgreyliteraturereports,togetherwithSwedisharchaeologicalreports.ADSdatasetsincludedtwoshipwreckdatasets,theNewportMedievalShip18andtheMysteryWreckProject(FlowerofUgie)19,togetherwiththeVernacularArchitectureGroupdendrochronology20 andCruck21 databases.DANS facilitated an extract from thedatabaseof theinternationalDigitalCollaboratoryforCulturalDendrochronology(DCCD22).

TheCIDOCCRMwasusedtoconnectthedataelementsandtheNLPentities,whichincludeObject,Sample,Material,Place(insomecases),dateranges.AspinevocabularywasidentifiedfromtheAAThierarchies forMaterial andObjects.Mappings fromDutch terms to AAT conceptsmostly existedalready from WP15 mapping work, while mappings from Swedish terms to AAT concepts wereproducedbySND,aspartoftheirWP16effort.

The NLP focus was on concepts relevant to the theme, such as samples, materials, objects andtemporal information. TheNLP techniqueswereable to generateXMLoutput fromEnglish,Dutchand Swedish texts, which was then transformed to the same RDF format as the instance data

17ForadiscussionofotherARIADNEcasestudiesofitemlevelintegration,seeARIADNED14.2Pilot

DeploymentExperiments18http://dx.doi.org/10.5284/102089819http://dx.doi.org/10.5284/101189920http://dx.doi.org/10.5284/103945421http://dx.doi.org/10.5284/103149722http://dendro.dans.knaw.nl

Page 14: D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

ARIADNE–D15.3:ReportonSemanticAnnotationandLinking PreparedbyUSW

ARIADNE 14 January2017

extracted andmapped to theCRM/AAT.NLPderivedRDF statements donot necessarily carry thesamedegreeofreliabilityasthosederivedfromthedatasets;theDutchandSwedishNLPpipelinesare at a prototype stage. The NLP outcomes include some false positives. See ARIADNE D16.4(SecondReportonNaturalLanguageProcessing)formoredetailsontheNLPtechniques,adiscussionof limitationsandfurtherwork. Infuturework,someindicationoftheprovenanceoftheRDFdatacouldbeincludedintheCRMmodel,whichwouldallowinformedjudgmentsofthereliabilityoftheinformation.

DatacleaningwasnecessaryandachievedviaOpenRefine23.TheoutputfromGATEwasexpressedasXMLandwas converted toRDF conforming to theCIDOCCRMwith connectionsalsomade to theAAT. This was achieved bymeans of a mapping/extraction tool, STELETO, developed by USW forARIADNE and freely available as open source24. STELETO is a 'lite' cross platform version of theSTELLAR.CONSOLE application developed for the STELLAR Project25 (Binding et al. 2015). It is ageneraldelimitedtextdataconversionutility,withsimplercommandlinefunctionalityandefficientperformance.STELETOconvertstabularinputdatatoanytextualoutputformatviaacustom(user-defined)template. ItwasusedbyUSWtoconvertthecasestudydata(frombothdatasetandtextstreams)sothatitconformedtothesemanticframeworkoftheCIDOCCRMandAAT.Intotal,1.09millionRDFtriplesresultedfrom23,594recordswith37,935objects.

Figure4:Demonstratorarchitectureandworkflow

The overall architecture and workflow is shown in Figure 4. The research Demonstrator crosssearchesthedataextractedfromdatasetsandDutch,EnglishandSwedishtextreportsviaSPARQLqueries. The Demonstrator is a SPARQL query builder, developed by USW, that seeks to hide the

23OpenRefine,http://openrefine.org24STELETO,https://github.com/cbinding/STELETO25STELLARProject,http://hypermedia.research.southwales.ac.uk/kos/stellar/

Page 15: D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

ARIADNE–D15.3:ReportonSemanticAnnotationandLinking PreparedbyUSW

ARIADNE 15 January2017

complexity of the underlying ontology. This Web application demonstrates the potential foralternative user interfaces to a plain SPARQL endpoint building on techniques developed in theSENESCHAL project26. As the user selects from the interface, an underlying SPARQL query isautomaticallyconstructedintermsofthecorrespondingontologicalentities.Itispossibletosearchacrossalldatasets(thedefault)orselectadatasettosearchindividually.Asetofinteractivecontrolsoffer search and browsing of the extracted archaeological data. The controls are designed to bebrowseragnosticandtheDemonstratorwillruninmostmoderninternetbrowsers.

Figure5:Dataintegrationcasestudydemonstratorapplication

Thedemonstratorcanperformsemanticallystructuredqueries, free-textqueries,oracombinationof both.Drop-down lists of all datasets, AATmaterials andAAT object types used in the data arepopulatedatstartup,andadualslidercontrolisinitializedtorepresenttheminimumandmaximumyearsforanyobjectproductiondatespresentinthedata.Thisprovidesusefulselectableoptionstoassistqueryformulation.HierarchicalexpansionhasbeenimplementedoverthesemanticstructureoftheGettyAATandresultsfromnarrowerconceptsareincludedwhenavailable.

Figure 5 illustrates the case study query builder application, performing a structured query onrecords linked to the AAT concept “Salix (genus)”. The multilingual results originate from Dutch,Swedish and English records, generated from databases and textual reports. Some of the resultsshownarereferringtotheAATconcept“willow”–this isbecausetheapplication is leveragingthe

26SENESCHALProject,http://hypermedia.research.southwales.ac.uk/kos/SENESCHAL/

Page 16: D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

ARIADNE–D15.3:ReportonSemanticAnnotationandLinking PreparedbyUSW

ARIADNE 16 January2017

underlying AAT structure to automatically expand the original query across relevant semanticrelationships,soimprovingrecallwithoutsacrificingprecision.

Figure6showsaqueryonobjectsdeclaredasbeingoftype“roofs”,withaproductiondatestatedasbeingsomewherewithinthedaterange1500to1600AD.TheresultsoriginatefromdataderivedviaNLPonatextualreport,indicatingsomeinstanceswheretheprocesshasassociatedanobjectwithadate(oradaterange).ExaminationofthetextualnotesassociatedwitheachobjectsuggestthattheNLPresultsarecorrect.

Figure6:Searchonspecificobjecttypeanddaterange

The Demonstrator is able to cross search and browse information extracted from datasets andreports in different languages via the common framework based on the CIDOC CRM and AAT; adegree of semantic integration between data extracted from text documents and databases hasbeenachieved.Infuturework,theintentionistoexplorepossibilitiesofthesemantictechniquesforlarger scale efforts onmultilingual integration of datasetswith reports. Another aim is to explorepossibilities for more intuitive user interfaces for searching RDF datasets than the usual SPARQLendpoint.TheDemonstratorisavailableforuseviatheARIADNEPortalservices27.ItisalsoavailableasopensourcewiththecodefreelyavailableonGitHub28.

27Wood/DendrochronologyDemonstrator,http://portal.ariadne-infrastructure.eu/services28Wood/DendrochronologyDemonstratorcode,https://github.com/cbinding/ARIADNE-data-integration

Page 17: D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

ARIADNE–D15.3:ReportonSemanticAnnotationandLinking PreparedbyUSW

ARIADNE 17 January2017

3 Summaryandlessonslearned

Briefsummary

The review and development work described in this chapter focuses on the aspects of semanticlinking and annotation particularly relevant to ARIADNE. The nature of semantic linking andannotationwithintheARADNEcontextisdiscussedandreferenceismadetootherrelevantARIADNEdeliverables.SemanticlinkingwithinARIADNEisconsideredwithinthespatial,temporalandsubjectdimensions. The subject dimension is considered in depth, starting with a review of linking toolsconsidered relevant to ARIADNE followed by a discussion of the ARIADNE approach and thevocabularymapping tools usedwithinARIADNE. TheGettyAATproved an appropriate vocabularymappinghub thataffordedamultilingual searchcapability in theARIADNEPortalvia thesemanticenrichment of partner subject metadata with derived AAT concepts. The characteristics of thedifferenttypesofmappingsofpartnervocabulariestotheGettyAATproducedwithinARIADNEaredescribed. In total 6416mappings were conducted, withmappings by individual partners rangingfromafewtoover1600terms.ByDecember2016,conceptsfrom27vocabulariesemployedby12project partners have been mapped to the AAT. The vocabulary mapping tools developed forARIADNEareavailableasopensourceviaGithub.

A case study conducted an exploratory investigation of the semantic integration of extracts fromarchaeological datasets with information extracted via NLP across different languages. Theinvestigationfollowedabroadthemerelatingtowoodenmaterialincludingshipwrecks,withafocusontypesofwoodenmaterial,samplestaken,woodenobjectswithdatingfromdendrochronologicalanalysis, etc. The Demonstrator developed for this wood/dendrochronological case study isdescribedwithillustrativescreendumps;itisavailableforgeneraluse.Theuserisshieldedfromthecomplexityoftheunderlyingsemanticframework(basedontheCIDOCCRMandGettyAAT)bytheWeb application user interface. The Demonstrator highlights the potential for archaeologicalresearch that can interrogategrey literature reports in conjunctionwithdatasets.Queries concernwooden objects (e.g. samples of beech wood keels), optionally from a given date range, withautomaticexpansionoverhierar-chiesofwoodtypes.

Lessonslearned

o The spatial, temporal and subjectdimensions are key to archaeology and themain vehicle forsemantic linking. Cleansing and normalisation via semantic enrichment is necessary in eachdimension.

o Inthesubjectdimension,ARIADNErequireshighqualityvocabularymappings,whichcannotbeachieved by automatic means alone, although appropriate mapping tools can help. It isimportanttosupportexpertintellectualreviewandprovidecontextualdata.

o The Getty AAT proved an appropriate vocabulary mapping hub that afforded a multilingualsearchcapabilityintheARIADNEPortal.

o TheWood/Dendrochronologycasestudydemonstratedthefeasibilityofconnectinginformation(at a detailed level) extracted fromdatasets and grey literature reports in different languagesand semantic cross-searching of the integrated information. The case study suggests that thesemantic linking of textual reports and datasets opens up possibilities for integrativearchaeologicalresearchacrossdiverseresources.

Page 18: D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

ARIADNE–D15.3:ReportonSemanticAnnotationandLinking PreparedbyUSW

ARIADNE 18 January2017

o Data cleansing is necessary for semantic data integration together with tools that ensure aconsistentrepresentationoftheextracteddataintermsofthesemanticframework.

o NLPderivedRDFstatementsdonotcarrythesamedegreeofreliabilityasthosederivedfromthedatasetsandNLPoutcomesmayincludesomefalsepositives.AnindicationoftheprovenanceoftheRDFdatashouldbeincludedinthesemanticframework,whichwouldallowjudgmentsofthereliabilityoftheinformation.

o The Wood/Dendrochronology demonstrator shows that a Web application can hide thecomplexity of the underlying semantic framework from the user and allow querying andbrowsing the information without expertise in SPARQL; it illustrates that more intuitive userinterfacesarepossibleforsearchingRDFdatasetsthantheusualSPARQLendpointorbrowsinghyperlinks.

Page 19: D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

ARIADNE–D15.3:ReportonSemanticAnnotationandLinking PreparedbyUSW

ARIADNE 19 January2017

4 References

AMALGAME,http://semanticweb.cs.vu.nl/amalgame/

ARIADNE(2015)website:ARIADNEatLinkedPasts,http://www.ariadne-infrastructure.eu/News/ARIADNE-at-Linked-Pasts

BindingC.&TudhopeD.(2016):ImprovingInteroperabilityusingVocabularyLinkedData.In:InternationalJournalonDigitalLibraries,17(1),5-21[doi:10.1007/s00799-015-0166-y]

CaraccioloC.,StellatoA.,MorshedA.,JohannsenG.,RajbahndariS.,JaquesY.&KeizerJ.(2013):TheAGROVOCLinkedDataset.In:SemanticWeb.4(3):341-348.

CaraccioloC.,StellatoA.,RajbahndariS.,MorshedA.,JohannsenG.,JaquesY.&KeizerJ.(2012):Thesaurusmaintenance,alignmentandpublicationaslinkeddata:theAGROVOCusecase.In:InternationalJournalofMetadata,SemanticsandOntologies,7(1):65-75.

CharlesV.,FreireN.&IsaacA.(2014):Links,languagesandsemantics:linkeddataapproachesintheEuropeanLibraryandEuropeana.ProceedingsIFLAWorldLibraryandInformationCongress2014,Lyon,http://ifla2014-satdata.bnf.fr/pdf/iflalld2014_submission_Charles_Freire_Isaac.pdf

CULTUURLINK,http://cultuurlink.beeldengeluid.nl/app/#/start

HeritageData.LinkedDataVocabulariesforCulturalHeritage,http://www.heritagedata.org

IsaacA.,WaitesW.,YoungJ.&ZengM.(eds.)(2011):LibraryLinkedDataIncubatorGroup:Datasets,valuevocabularies,andmetadataelementsets[W3CIncubatorGroupReport,October25,2011].http://www.w3.org/2005/Incubator/lld/XGR-lld-vocabdataset/

ISO25964-2:2013.Informationanddocumentation-Thesauriandinteroperabilitywithothervocabularies-Part2:Interoperabilitywithothervocabularies.http://www.niso.org/schemas/iso25964/#part2

ISO25964-1:2011.Informationanddocumentation-Thesauriandinteroperabilitywithothervocabularies-Part1:Thesauriforinformationretrieval.http://www.niso.org/schemas/iso25964/#part1

KempfA.,NeubertJ.,FadenM.(2015):TheMissingLink–AVocabularyMappingEffortinEconomics.14thEuropeanNetworkedKnowledgeOrganizationSystems(NKOS)Workshop,TPDL2015Poznań.

LiangA.&SiniM.(2006):MappingAGROVOCandtheChineseAgriculturalThesaurus:Definitions,tools,procedures.In:NewReviewofHypermediaandMultimedia,12(1):51-62.

MeghiniC.etal.(forthcoming):ARIADNE:AResearchInfrastructureforArchaeology.In:JournalonComputingandCulturalHeritage.

MilesA.&BechoferS.(2009):SKOSSimpleKnowledgeOrganizationSystemReference.W3CRecommendation,https://www.w3.org/TR/2009/REC-skos-reference-20090818/#L4138

NiccolucciF.&HermonS.(2015):Representinggazetteersandperiodthesauriinfour-dimensionalspacetime.In:InternationalJournalonDigitalLibraries,17(1):63-69[doi:10.1007/s00799-015-0159-x]

OpenAnnotationOntology,http://www.openannotation.org

OpenRefine,http://openrefine.org

Page 20: D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

ARIADNE–D15.3:ReportonSemanticAnnotationandLinking PreparedbyUSW

ARIADNE 20 January2017

Pelagios,http://commons.pelagios.org

PeriodOclient,http://n2t.net/ark:/99152/p0

PeriodOwebsite,http://perio.do

PeriodO(Github),https://github.com/periodo

Peripleo,https://github.com/pelagios/peripleo#peripleo-api

Pleiades,https://pleiades.stoa.org

Recogito,http://recogito.pelagios.org

RichardsJ.,TudhopeD.&VlachidisA.(2015):TextMininginArchaeology:ExtractingInformationfromArchaeologicalReports,pp.240-254,in:BarceloJ.&BogdanovicI.(eds.):MathematicsandArchaeology.CRCPress

SAIMInstanceMatchingApplication,http://saim.aksw.org

SENESCHALproject:SemanticENrichmentEnablingSustainabilityofarCHAeologicalLinks.UniversityofSouthWales,HypermediaResearchGroup,http://hypermedia.research.southwales.ac.uk/kos/seneschal/

ShawR.,RabinowitzA.,GoldenP.&KansaE.(2015):Asharing-orienteddesignstrategyforNetworkedKnowledgeOrganizationSystems.InternationalJournalonLibraries,17(1):49-61[doi:10.1007/s00799-015-0164-0]

ShawR.,RabinowitzA.,GoldenP.&KansaE.(2015):ReportonanddemonstrationofthePeriodOperiodgazetteer.14thEuropeanNetworkedKnowledgeOrganizationSystems(NKOS)Workshop,TPDL2015Poznań.

SILK-SilkLinkDiscoveryFramework,http://wifo5-03.informatik.uni-mannheim.de/bizer/silk/

SimonR.,IsaksenL.,BarkerE.&deSotoCañamaresP.(2016):Peripleo:aToolforExploringHeterogenousDatathroughtheDimensionsofSpaceandTime.In:Code4LibIssue31,2016-01-28,http://journal.code4lib.org/articles/11144

SimonR.,IsaksenL.,BarkerE.&deSotoCañamaresP.(forthcoming):ThePleiadesGazetteerandthePelagiosProject,in:BermanM.,MosternR.&SouthallH.(eds.):PlacingNames:EnrichingandIntegratingGazetteers.IndianaUniversityPress.

SKOSmappingrelationships,http://www.w3.org/TR/skos-reference/#L4138

SKOS.SimpleKnowledgeOrganizationSystem.W3C,http://www.w3.org/2004/02/skos/

SPARQL1.1QueryLanguage.W3C.(2013),http://www.w3.org/TR/sparql11-query/

StahnL.(2015).VocabularyAlignmentforarchaeologicalKnowledgeOrganizationSystems.14thWorkshoponNetworkedKnowledgeOrganizationSystems,TPDL2015Poznań.

STARProject:SemanticTechnologiesforArchaeologicalResources.UniversityofSouthWales:HypermediaResearchGroup,http://hypermedia.research.southwales.ac.uk/kos/star/

STELETO,http://github.com/cbinding/STELETO/

STELLARProject:SemanticTechnologiesEnhancingLinksandLinkedDataforArchaeologicalResources.UniversityofSouthWales:HypermediaResearchGroup,http://hypermedia.research.southwales.ac.uk/kos/STELLAR/

Page 21: D15.3 Report on Semantic Annotation and Linking - ARIADNE PLUSlegacy.ariadne-infrastructure.eu/wp-content/uploads/2019/07/ARIAD… · ARIADNE D15.1 gives an overview of vocabulary

ARIADNE–D15.3:ReportonSemanticAnnotationandLinking PreparedbyUSW

ARIADNE 21 January2017

StillerJ.,PetrasV.,GädeM.&IsaacA.(2014):AutomaticEnrichmentswithControlledVocabulariesinEuropeana:ChallengesandConsequences,pp.238-247,in:DigitalHeritage.ProgressinCulturalHeritage:Documentation,Preservation,andProtection.SpringerLNCS8740.

ZengM.&ChanL(2004):Trendsandissuesinestablishinginteroperabilityamongknowledgeorganizationsystems.In:JournalofAmericanSocietyforInformationScienceandTechnology,55(5):377-395.