Post on 14-Mar-2018
Kobe University Repository : Kernel
タイトルTit le
COLLABORATIVE KNOWLEDGE MAP FOR DISASTERMANAGEMENT COLLECTIONS
著者Author(s) Parmanto, Bambang
掲載誌・巻号・ページCitat ion 神戸大学都市安全研究センター研究報告,13:113-133
刊行日Issue date 2009-03
資源タイプResource Type Departmental Bullet in Paper / 紀要論文
版区分Resource Version publisher
権利Rights
DOI
URL http://www.lib.kobe-u.ac.jp/handle_kernel/81001958
Create Date: 2018-05-17
神戸大学都市安全研究センター研究報告,第13号,平成21年 3 月
― ��� ―
ColbborativeKnowledgeMapfわrDisasterManagementCollections
BambangParmanto,UniversltyOfPittsburghandRobeUniverslty
1.Introduction
lnpreparlngOrrespondingtodisasterevents,profTessionalswhoworkindisaster
managementusuallysearchfわrcasesfromthepastthatcloselymatchedwiththecurrenteventas
arefTerence.Findingandretrievingdocumentsfromthepasteventshavebeendifficulteven
whendatabasescontainingdigitallibrarycollectionsofthepasteventsareavailable.Thereare
manychallengesthatweneedtoaddressifwewanttomakethedisastercollectionsmore
reusableforresearchersandprofessionalswhoworkindisastermanagement.Thefirstchallenge
isthefactthatdisastercollectionsarescatteredinvariousinstitutionsaroundtheworld.
Numerousorganizationsinvolvedinthedisastermanagementlocatedinnumerouscountries
aroundtheworldmaintaindisastercollections,o洗ennotf♭rthemosteffTectiveuseandf♭r
indeterminateamountoftime.Disastercollections,consistlngOrdataandvarioustypesof
documents,areisolatedinthesesilosandinaccessibletooutsideresearchersandindividualswho
workindisastermanagement.
Thesecondchallengeisthemultilingualnatureorthecollections.Disastersknowno
languageandpoliticalboundaries.Forexample,theGreatSumatraTsunamidisasterof26
December2004affTectedmostcountriesintheIndianOceanfromIndonesiatoMadagascar.
Documentsintensornationallanguagesandevenmorelocaldialectsorlanguageshavebeen
producedfTromthisevent.ThethirdproblemistheheterogeneltyOfthecollections,especially
heterogeneltylnSemanticandtypesofcollections.Disastermanagementismultifacetedinits
natureanditsrangeofimpactandhencerequlreSmultidisciplinaryapproach.ProfTessionalswho
workindisasterpreparednessandresponsearedrawnfromsuchdiversebackgroundsas
sociology,englneerlng,healthsciences,andpublicadministration.Findingcommongroundwith
respecttothelanguagethatisusedtodenneandarticulatethedisasterfieldhasbeenachallenge
(Britton,1999).AvastmajorityofthedisastercollectionsarethesoICalled"greycollections":
― ��� ―
collectionsthatarenotfわrmallypublished(Lehman皮Webster,2004a).Thesecollections
includepamphlets,disastereducationmaterials,1nStitutionalreports,personalplCtureS,etC.
PeoplerememberpastdisasterfrompICtureSandfilmsthatarenotformallypublished.
Withoutproperinfわrmationmanagement,Weriskloslngtheinvaluablelessonslearned
fromthiscatastrophiceventandfromtherecoveryeffortsfromthepastdisaster.Thereuseof
infわrmationthatisscatteredinvariousorganizationsallovertheworldisveryImportant.Since
disasterdataandinfわrmationareownedbyvariousorganizationsallovertheworld,single
repositorylSnotanOPt10n.However,aunifiedadvancedmetadatathatallowmoreefficient
organizationandaccesstothedisasterdataandinfわrmationisneeded.Therefわre,theoptlmal
solutionwouldbefTederatedrepositorieswithcollaborativeadvancedmetadata.
Thegoalofthepaperistoproposethedevelopmentofacollaborative,unified,and
advancedmetadatafわrdisastercollectionsintheworld,especiallyinAsia.Theadvanced
metadatawillbecalled"knowledgemap"inthispaper.Theknowledgemapwillbeusedby
collaboratlngCOuntriesandinstitutionstomanagetheirowncollectionsandtoexchange
infわrmationbetweentheinstitutions.TheknowledgemapISCOnCePtualizedasaunifiedviewof
thefederativecollectionswiththegoalofmakingthecollectionsaccessibletousers.The
advancedmetadatawillhavethefollowingrequirements:Capableofrepresentlngunified
knowledgeofdisastermanagementfield,canbeusedtoimprovetheprecisionandrecallofthe
disastercollectioninfわrmationretrieval,Caneasilysupportmultilingualaccessandcollections,
canbridgemultidisciplinaryviewofthefield,CanbeusedinafTederatedenvironments,andtake
advantageoftheadvancesinInternettechnologies.
AnewstandardandasetoftechnologiescalledTopicMap(Pepper,2002)haverecently
beendevelopedandadoptedasanISO(InternationalStandardizationOrganization)standard.
TopicMapmeetsalltherequlrementSf♭rthecollaborativemetadataandknowledge
managementfわrdisastermanagementasdiscussedinthepreviousparagraph.Thispaperwill
fTocusonthepotentialofTopicMapfTorthedevelopmentofcollaborativeknowledgemapfわr
disastermanagement.Inadditiontomakingdisastercollectionreusableandfindable,Topic
M叩 mayalsoleadtoimprovementofourknowledgeindisasterinthesamewaytaxonomywas
usedtoimproveourknowledgeofthedomain(Kreps,1989;Quarantelli,1987).Knowledgemap
― ��� ―
issomethingthatcollectsthekeyconceptsinthedisastermanagementfieldandtiesitall
together,
2.DisasterCycle:InformationM anagementPerspective
Thefieldofdisastermanagementusuallydividesthetimeframeintofourphases:
mltlgation,preparedness,response,andrecovery.Thesefわurphasessometimesarealsodivided
intofわursimplephases:befわre,during,anda洗er.Thegoaloftheeffortbeforethedisasteristo
reduceoravoidlossesfromhazards,whilethegoalofeffortsduringdisasteristoassureprompt
assistancetotheaffTectedcommunlty,anda洗erthedisasterthegoalistoachieverapidand
effectiverecovery.Themltlgationandpreparednessphasescorrespondtothephasebefわrethe
disaster,theresponsephasecorrespondstoduring,andrecoverycorrespondstothephaseafter
thedisaster.TheMitigationPhaseisconcernedwithminimlZlngtheeffTectsofpossibledisasters
bymandatingpolicies(suchasbuildingcodesandzoning),conductingvulnerabilityanalyses
andpubliceducation.ThePreparednessPhaseconsistsofplannlngOnhowtorespondtoa
disaster,includingemergencyexerciseandtralnlng;andearlywarnlngSystem.TheResponse
phasedealswiththeaRermathofthedisasterandincludessuchactivitiesassearchandrescue,
andemergencyrelief.ThefinalphaseofRecoveryInvolvestheeffortsofretumlngthe
communltytOnormal,includingtemporaryhouslngandmedicalcare.
Duringpre-crisisphases,researchersandprofessionalsconductingdisastermltlgationand
preparednesscanlearnfromdisastersinthepastorfromsimilardisastersinotherregions.For
example,infわrmationsuchashazardanalysis,buildingcodes,andpubliceducationmaterialscan
bereused.Previousdisasterscouldalsoprovideinfわrmationwhichmethodsdidanddidnot
work.Plannlngmethodinoneplaceorfromthepastcouldbeadoptedormodifiedtosuit
disasterpreparedness.DuringcrlSIS,infわrmationexchangebetweenorganizationsinvolvedin
theresponseandreliefeffTortsismandatorytoensuretheeffectivenessoftheeffTorts.
Inthepostcrisisphases,collectinglessonslearnedfromthedisastershouldtakeplace.
Fromthelessonslearnedfromthecrisis,methodsortoolsthatarenolongervalidcanbe
adjusted.ThefわcusoftheproposedknowledgemapISinorganizlnglessonslearnedfromthe
pastdisasters,emphasizinginfわrmationreuseandinfわrmationexchange.Thelongtermgoalis
― ��� ―
toprovideinfわrmationthatiscurrentandtopassontheknowledgegainedfromthepastdisasters
fTorbetterpre-Crisismltlgationandpreparednessefbrts,andbetterreliefandresponseeffTorts.
PhasesinadisasterarealsorefTerredtoasdisastercycle.Fromtheinformation
managementperspective,disastercycleposesaunlqueChallenge.Unlikeothercyclethat
repeatedlyhappensinthesameplace,manytypesofdisaster(earthquake,tsunami,typhoon)
rarelyhappensrepeatedlyexactlyinthesameplace.Diverseinfわrmationandknowledgeare
widelydistributedandownedbyvariousorganizationslocatedindiffTerentcountries.Therefわre,
thelessonslearnedfromadisasterinoneplacemightgetlossbeforeotherplacesintheworld
couldleanfromthem.
AnothercharacteristicofdisastercycleisthelongtlmeSpananduncertaintybetween
pre-disasterphases(mitigationandpreparedness)andthedisaster.Formanytypesofdisaster,
weneverknowwhenadisasterwillstrike.Theperiodofuncertaintyisthenfわllowedbyfrantic
activitiesduringthedisasterresponsephase.Duringthisphase,allinfわrmationneededfわrrelief
activitieshastobeavailableinaninstant.
3.ReviewofRelatedWork
Twoareasarerelevanttothiswork:greycollectionofdisasterarchive,anddisaster
metadataandontology.Recentworkinthesetwoareaswillbediscussed,aswellasthe
relationshipbetweentheproposedknowledgemapwiththeserelatedworks.
3.1GreyCollectionofDisasterArchive
ThemaJOrltyOfdisastercoHectioncanbecategorizedasthesoICalled"greyliterature".
Greyliteratureiscommonlydefinedasanydocumentarymaterialthatisnotcommercially
publishedandistyplCallycomposedoftechnicalreports,workingpapers,businessdocuments,
andconferenceproceedings(Mathews,2004).Inthecaseofdisastercollection,typical
collectionsincludeone-pagedocument(suchaspamphlet),reports,newspaper,magazines,
reports,proJeCtdeliverables,spatialdigitaldatasets,data-setsderivedfromfieldwork,tralnlng
anddisseminationmaterials,andimagesorvideos.
Disastercollectionsareo洗enlocatedinvariousorganizationsallovertheworld.Manyof
thesearestoredinthewebsiteorcomputeroftheseorganizations.Oneofthechallengesisthe
― ��� ―
longevltyOfthecollectionisusuallynotasstableasthecollectionsmaintainedbylibrariesor
universities.Forexample,Oncethedisasterrecoveryandreconstructionworkiscompleted,the
maintenanceandsupportofthedisastercollectiongraduallyfade.Thesecollectionsarealsonot
maintainedfわrpromotlnginfわrmationsharingbetweenresearchers,Communlty,and
profTessionalsworkingindisastermanagement.
Sincedisasterinfわrmationanddatausuallyarealreadybeencollectedbyinstitutionsand
havebeenidentifiedbysubjectexpertorqualifiedlibrariansasrelevanttotheirwork,
identificationandacqulSlt10nusuallyarenotthegreatestchallengeasisthecasewithgrey
literature.Therefわre,thetermcollectionwillbeusedtorefTertomaterialsthathavebeen
identifiedasuseful.However,editorialcontrolandmetadatawillstillbeachallengeinthis
federatedenvironment.
ArelatedworkinthisareaisthedevelopmentofTsunamiDigitalLibrary(TDL)atthe
UniversityofOregon(LehmanandWebster,2007).Theyconductedsurveyandinterviewof
expertswithinthetsunamicommunity(LehmanandWebster,2007)andfわundthatthe
communltylSaSlgnificantportionofthematerialsrequiredanddesiredbytheinterdisciplinary
tsunamicommunltylSgreyandthesearenotadequatelyhoused,maintained,ororganizedby
existlngPartnerSites.Thestudyalsofわundthattheinfわrmationretrievalresults什omthecurrent
digitallibrarieshavebeenpoor.
3.2DisasterMetadataandOntology
ThestudyofdisasterrequlreSmultidisciplinaryapproachdotoitsmultifacetednatureand
rangeofimpacts.Thenumberofinstitutionsandcountriesinvolvedinthedisastermanagement
alsoaddtothediversltyOftheapproachtodisaster.TheresultisnotonlyintheheterogeneltyOf
theinfわrmationsystemsusedfわrmanaglnginfわrmation,butalsointheheterogeneltyOfthe
semantic.Peopleinthedomainhavenotfわundcommongroundwithrespecttothelanguagethat
isusedtodefineandarticulatethefield(Britton,1999).Thisproblemimpedesmany
interdisciplinarydevelopments,andcausesconfusionanddistancingbetweentheresearchersand
thepractitioners.TherehavebeenanumberofeffTortsthataddresstheneedf♭runified
conceptualdescrlPt10nOfthefield.Theeffortsrangefromasimpletaxonomytoformalontology.
― ��� ―
OntologylSanexplicitspecificationallowstheconceptsandrelationshipsinthemodeltobe
sharedbetweenpeople,awayf♭racommunltytOagreeOnCOmmOntermsf♭rcapturlngmeanlng
orrepresentingknowledgeinsomedomain(Uschold皮Gruninger,2004).Thegoalofthese
effortsissemanticinteroperabilityamongheterogeneousinformationsystems.
CurrenteffTortintaxonomicandontologlCalapproachestodisastershaveconcentrated
almostexclusivelyontheResponsePhase,dealingwiththeimmediatea氏ermathofan
emergencyIncludingthemobilizationofreliefagencies,thedeliveryofaidandtheprovisionof
emergencymedicalcare(Halliday,2007).Theworkindisasterontologycanbedividedintotwo
sources:USgovernmentagenciesthattrytomakeinfわrmationsystemsusedindisaster
managementtalktoeachother,andacademicexercisefromanumberofuniversltyresearch
groups.AnexamplefromthefirstgroupistheexistlngStandardsaretheCommonAlertlng
Protocol(CAP)andtheEmergencyDataExchangeLanguage(EDXL).Thestandardfわcusesthe
communicationbetweenemergencyresponseinfわrmationsystems,especiallywithintheUS
govemmentalinstitutions.Thereisalsolimitedevidenceastoitssuccessfuladopt10n
implementationandapplicationsinrecentemergencies(DiMaio,2007).Thesecondgroup
focusesonthedevelopmentofmoregeneraldisastermanagementontologywiththegoalof
achievlnglnterOperabilitybetweenvariousinfわrmationsystemsusedinthedisastermanagement
(Sotoodeh&Kruchten,2008;Joshieta1.,2007).Theeffortsindisasterontologyareprimarily
devotedtoachievesemanticinteroperabilitybetweeninfわrmationsystemsusedindisaster
management.TheseontologleS,aSabridgebetweenheterogeneousdatasourcesandinfわrmation
systems,areusedtosupportinfわrmationbrokeringandmediation.Thereisnoevidencethatthe
worksindisasterontologyhasbeenusefulbeyondacademicexerciseforontologydevelopment.
ItislikelythattheontologylStooCOmPlicatedandtooinnexibletobeusefulorproductive.
Minimalworkhasbeendoneinthedevelopmentofontologywiththegoalofincreaslng
thefindabilityandorganlZlngCOllectionsfromthepastdisasters.However,thebodyof
knowledgeaccumulatedfromtheeffTortindisasterontologydevelopmentcanbeusedinthe
developmentofDisasterKnowledgeMapasproposedinthisp叩er.
4.KnowledgeMapofDisasterManagement
4.1KnowledgeManagementofDisasterCollections
― ��� ―
Thepu叩OSeOfaKnowledgeManagement(KM)indisastermanagementistosupport
decisionmakingbasedoninformationandexperienceintheregionandfrompastdisasters.
KnowledgeanddatafTromthepastdisastersareinvaluableresourcesfわrmltlgatlng,PreParlng,
andrespondingtofuturedisasters.Whenworkingonanyphaseofdisaster,professionalsusually
searchforcasesthatbestmatchthecurrenteventfromthehistoricalcasedatabasesasa
reference.Thegoaloftheproposedadvancedmetadataistosupport"resourcediscovery"or
infわrmationretrievalofthecollectionsrelevanttodisastermanagement(especiallycollections
frompastdisasters),aswellastosupportthemanagementofthecontentsofnewacquisition.
lnthispaperwewillusethetermcollectiontorefTertoallobjectsthatneedtobe
organized.ThecollectionsconsistofdiscretepleCeSOfobjectthatwillbecalleddocument.The
termdocumentwillbeusednotinthetraditionalsenseortheword.Documentcanbeinthefわrm
orpaper,book,image,video,sound,numericaldata,andmap.Thedisastercollectionsconsistof
varioustypesordocumentsrelevanttodisastermanagement.
4.2Metadata,Taxonomy,andEndl0f-the-bookIndex
ThemostcommonmethodforaidinginfわrmationretrievalfromtherepositorylSuSlng
metadata.AnexampleofmetadatafordisastercollectionoftheGreatHanshin-Awaji
EarthquakeDisasterMaterialCollection(EDMC)atRobeUniversitycontainsthefわllowing
items:title,author,publisher,size,place,Callnumber,typeofmaterial/collection,and
publicationdate.Aswecansee,thestandardtraditionalmetadatacontainsmostlyadministrative
infわrmation.Thereisnotmuchinfわrmationaboutthecontentofthecollectionsthatcanhelp
userretrieveadocumentaccurately.Theonlywayfわrtheuserstosearchthecollectionis
throughmatchingthewordwiththetitle,authorname,orpublicationdate.Traditionalmetadata
doesnothelptheusersinjudgingthecontentsofthecollectionsbeyondthetitle.Somemay
includeabstract,butthatmeanstheusersneedtoreadtheabstractonebyonetodecidewhich
onesarerelevant.Traditionalmetadatausuallyleadstoimprecisesearchandpoorinfわrmation
retrievalaccuracy.
Inadditiontothestandardmetadata,anothertoolthatiscommonlyusedtoaidretrievalis
controlledvocabularyortaxonomy.Taxonomyinthisproposalisreferredoasasubject-based
classificationthatarrangesthetermsinthesubjectcontrolledvocabularyIntoaStatichierarchy.
― ��0 ―
Forexample,theRobeUniverslty'SEDMCcontainssimpletaxonomyconsistlngOfl6items:
earthquakedisaster,law,economy,publicadministration,flreService/disasterprevention,citizen
lives,socialwelfare,education,earthquakeenglneering/groundvibration,medicalcare,civil
englneering/architecture,industry/physicaldistribution,traffic/harbor,
infわrmation/communications,art/culturalheritage,andliterature(personalstory,literarywork,
etc.).Whilestandardmetadataonlylinksthemetadataitemstothecollection(object),taxonomy
arrangesthecollectionintoahierarchy.TaxonomylSnotpartOrthemetadata,althoughitcanbe
usedinmetadata.ThebenefitofaddingmoredescrlPt10ntOthetraditionalmetadataisricher
descrlPt10nOfthecollectionwiththeultimateobjectiveofmoreprecisesearchandretrieval.
Anothermethodthathasbeenusedtoimprovethefindabilityofdocumentsina
collectionistheuseoffull-textindex.Insteadofhavingaccessonlytotitleorabstractjull-text
indexprovidesaccesstothewholecontentsthroughthewordsbeingindexedfromthecontents.
Theproblemwithinvertedindexisthatallwordsinthecontents,WiththeexceptlOnSOfstop
words(suchwordasand,or,andis),areindexedandareconsideredequallyimportantconcepts
inthedocument.RetrievalfTromfu111teXtindexedcollectionsyieldshighnoise(lowprecision)
sincealldocumentscontalnlngwordsthatmatchthesearchwordwillberetrieved.
Amorecleverwaytoachievlnghighprecisionoftheretrievalistheuseback-in-thebook
index.Theindexattheendofthebookisconstructedtorepresenttheknowledgeoftheentire
book.Endofbookindexisinfactamapoftheknowledgecontainedinabook:itcontainsthe
essenceandonlysalienttoplCSOfthebook.Endofbookindexprovidesmuchbetterprecision
forretrievalthanthefull-textindex.However,endofbookindexisonlyusefulforretrievlng
sectionsofabook,notacollectionsofthousandsbooksandothermaterials.
4.3DisasterKnowledgeMapUsingTopicMap
AnewstandardandasetoftechnologleSCalledTopicMapusessimilarmethodasthe
backofbookindex,butinamuchmorepowerRllway.Thisnewstandardhasbeenadoptedby
ISOandnumeroustechnologleStOimplementthestandardhavebeendevelopedinthepastfew
years.ThisnewtechnologylSagoodfitforknowledgemanagementofdisastercollections,fora
numberofreasons,prlmarilybecauseit:
― ��� ―
・ canbeusedtodevelopaknowledgemap,anadvancedversionofthebackofthebook
indexthatrepresentstheessenceofthedisastermanagementfield.
・ elegantlyhandlesmultilingualknowledgemapandmultilingualcollections.ThetopIC
mapISSeparatefromthecollection,bothofthemcouldbemultilingual
・ elegantlyhandlesmultimodaldocuments,fromimagestoJOurnalpapers.
・ CanrepresentcomplexknowledgerepresentationsthroughtoplCS,hierarchy,and
associations,handlesdiffTerentviewofusers.
・ CanaccommodatediffTerentviewsofmultidisciplinaryapproachtodisastermanagement.
ThefbllowlnglSabriefillustrationonhowtoplCmapCanbeusedtodevelopknowledge
managementordisastercollections.SincetheideaoftoplCmapCanbetracedbacktotheback-
of-the-bookindex,ltiseasiertoillustratehowitworksuslngexamplesoftheendbookindex.
Figure1illustratestheindexfわrahypotheticalbookontsunami.
Earthquake 12,34,35.231
Tektonik ll,2323
Volcanic 12,23.
Explosion 45,65.
Tsunami 121,
seealsoEarthquake
Hanshin-Awaji 12,23,34,445
Figure1.Asnippetofback-of-the-bookindexofahypotheticalbookontsunami.
― ��� ―
Themaincomponentsoftheback-of-thebookarenamesoftopics(earthquake,tsunami,
etc)andreferencetooccurrencesofthosetopics.Forexample,topic"earthquake"occursin
pages12,34,35,and231.1nadditiontoahierarchyortopics(earthquake:tectonicand
volcanic),anindexmayalsoincludeassociationsuchastheuseofsee,andseealso.Theuseof
seereferencesmakestheindexcapableofhandlingsynonymsormultipleentriesofnamesfor
thesametoplCOrconcept.Anindexmayalsousetypographicalconventiontodistinguish
diffTerenttypesoftopICSOroccurrences.Forexample,"Hanshin-Awaji"iswritteninboldto
showthatitisanameofadisasterevent,insteadofconcepts.Theoccurrenceinpage230ftoplC
volcaniciswrittenasredanditalictodistinguishthatthetoplC"VOIcanic"islinkedtoanm叩On
page23,insteadoftheword"volcanic"onpage23.
TopicmapISanadvancedcomputerizedversionofback10f-the-bookindex.Themain
constituentsofTopicMap(TM)arealsoTopics,Occurrences,andAssociations.Similartothe
index,TopicinTMrepresentssubjectsorconceptsofthedomainoftheTM.ThediffTerenceis
thatthesubjectinTMcoverstheentirecollections,insteadofjustonebook.Occurrenceisa
linkbetweentoplCSandthedocumentsinthecollection,similartopagenumberintheendbook
index.AtoplCCanbelinkedtomorethanonedocuments,andviceversa.
Figure2isasimpleillustrationofaTopicMapinadisastermanagement.Weborrow
thesameexampleofFigure1,partsofthetoplCSinourdisasterknowledgemapcontainstoplC
conceptsof"earthquake"thathassubtopicof"volcanic","tectonic",and"explosion."Italso
containstoplCCOnCePtSOf"tsunami"and"HansinAwaji."ThediffTerenceisthatthisconcept
willpointtoadocument(map,journalp叩er,etC)intherealdigitallibrarycollection.lnourcase,
WechoosetorepresentdocumenttypeasanOccurrence:occurrenceorjoumalpaperwilllinkall
Journalp叩erSrelevanttoeallhquaketothetoplCconceptearthquake.Occurrencewillprovide
morethanjustasimplelinksinceOccurrenceitselfcanbeahierarchyoranetworkofconcepts.
Forexample,anoccurrenceofperiodicalscanhesuboccurenceormagazine,Journalpaper,
newspaper.Wecanuseoccurrence"periodicals"toretrievedocumentrelevanttothetopICWe
choosethatareavailableinallfわrmatsunder"periodical"(magazine,journalpaper,and
newspaper).
OnecomponentthatmakesTopicMapverysophisticatedisthepowerofAssociatL'on.
Unlikelimitedassociationavailableintheindex(simplehierarchy,see,andseealso),
― ��� ―
AssociationinTopicMaplSagr叩hthatcanlinkanytoplCtOanotherwithanassociationthat
canbedefinedbythedesigner.Inotherwords,TMisextensible:TM istoendbookindexthe
wayXMListoHMTLInourexampletheAssociationbetweentsunamiandearthquakeis
deflnedas"causedby"(tsunamiiscausedbyanearthquake).Onetopiccanalsohavemultiple
supertopic(parenttopic).Forexample,theCityofKobecanbepartofahierarchyof
administrativeunitinJapan(Japan- HyogoPrefecture- CityofKobe),butcanalsobepartof
atectonicplatehierarchyof(Eurasianplate-Niigata-RobeZone- CityofKobe).TM
technologleSWOuldallowustodevelopanadvancedknowledgem叩thatrepresentstheessence
andallsalienceconceptsindisastermanagement.
AssociationisusedextensivelyinTopicstofわrmanetworkofconnectedconcepts.As
TMisextensible,AssociationcanalsobeusedfわrOccurrence.Forexample,ifweuselanguage
orthedocumentasoccurrence,Wecanhaveahierarchyorlanguageandassociationof
languages.Wecanalsousetypesofdocument(map,books,periodicals,etc)inamorecomplex
association.Forexample,reportswithextensiveuseofmapcanbecategorizedasbothmapand
report・
COLLECTIONS
- -p -A+ 0
-----0
m ヰ姦 ⊇
・OCCURENC毘
-Jnun]礼pBP_er_I._
vol占aniTL-- 丁~-----∀柑e9----___
TOPICS
Earthquak・e一一-一 一・一Tektomic I、\
l __ \
ExpI'osi・01, Cals如/
EilTstmami
Hansin-AwaJA
Figure2.SimpleillustrationoftoplCmapfわrdisastermanagement
― ��� ―
Topicm叩teChnologyallowstheuseofscopeinTopic,Occurrence,andAssociation.
Scopeissimilartothesaurusintraditionallibrary,onlymuchmorepowerful.Theuseofscope
canenableustodealwithdifferentterminologiesusedbyvariousdisciplinesindisaster
management.Forexample,highintensltytrOplCalstormiscalled"typhoon"inEastAsiareglOn
(Japan,Taiwan,toIndonesia)whileitiscalled"cyclone"inAustralia,andisreferredtoas
HhurricaneHintheAmericas.
ThebenefitofusingTopicMapapproachfordevelopingknowledgemapisthatitleads
tomorerichlydescrlPtlVeClassificationandmetadatasystemsthananyofthetraditional
approaches.Richclassificationandmetadatasystemswillleadtoinformationretrievalwithhigh
recallandprecision.Itwouldalsoallowf♭rthedevelopmentofnavlgationsystemsdrivenbythe
knowledgemaptobemuchricherandmorenexiblethanasimpletree-basednavlgation.
4.4TopicandTopicClassesofDisasterCollection
Knowledgemapfわrdisastermanagementwillbedevelopedbyconductlngextensive
interviewswithresearchersanddisastermanagementexperts.DeveloplngtOplCmapisakinto
developlngadvancedbackl0f-the-bookindexthatreflectsknowledgestructureofthedisaster
managementfield.DiffTerentunderstandingofconceptsbetweenvariousdisciplinescanbe
accommodatedinknowledgem叩,somethingthatcannotbedoneinanendbookindex.The
followlnglSabriefinitialthoughtonpotentialtopICClassesindisastermanagement.ThetopIC
classesindisastermanagementmayIncludethefbllowlngltemS:
・ Disastertaxonomy:naturaldisaster,humansystemfailure,connict-baseddisaster.
Naturaldisaster:earthquake,tsunami,mood,etc.Thistaxonomycanbedevelopedtothe
detailedlevelofspecificfieldsinthedisastermanagement.Wecancombineprevious
work(Green&McGinnis,2002;Kreps,1989,etc)andinterdisciplinaryinputfrom
variousexperttodevelopthetaxonomy.
・ Spatialorgeographicallocationthatwillallowinfわrmationretrievalbasedonlocation,
eitherlatitude-longitude,orbasedonspatialtaxonomysuchasadministrativehierarchy,
tectonicrelationships,etc.
・ TimedimensioninDisastermanagementcycle:Mitigation,preparedness,response,
recovery(reconstruction).Morennegrainedtimingcanalsobedevelopedtoallow
infbrmationretrievalbasedonthedisastercycletime.
― ��� ―
・ Bigdisasterevent.Manycollectionsanddataarerelatedtobigdisastereventssuchas
HurricaneKatrinaand9/11terroristattackintheUSA,Ilanshin-AwajiEarthquakein
Japan,SumatraTsunamidisasterinSouthandSoutheastAsia.
・ Finegrainedconceptsindisastermanagement,butmoregeneralthancurrentworkon
ontologyofdisastermanagement.Again,acombinationorcurrentworkondisaster
ontologyandinputfromdisastermanagementexpertscanbeusedforthedevelopmentof
theseconcepts.
・ Traditionalmetadata:time,title,author,subject,abstract,etc.
4.5MultilingualApproach
ThemultilingualrequlrementSf♭rthedisastermanagementcollectionshavetwo
dimensions:
・ Theknowledgemap(topicmap)needtobemultilingual,thinkofhavingadvancedback-
of-the-bookmultilingualindex
・ Thecollection(documents)canbeinmultiplelanguages.
Theknowledgem叩 needstobemultilingualirrespectiveorthelanguageofthe
documentsinthecollections.Thisknowledgemapcanbeusedasashareresourcebymany
organizationslocatedinvariouscountries.Usingthisapproach,auserwillbeabletosearchthe
collectionuslngthesearchlanguageoftheuserbutwillretrieverelevantdocumentinany
languages.
InordertosupportthemultilingualrequlrementS,tOplCClassesneedtobetranslatedin
variouslanguages.ThiscanbehandledeasilyinTopicMap.ThefbllowlnglSanexampleofan
entryintheTopicclassinthreelanguages(English,Japanese,andIndonesian/Malaysian):
― ��� ―
TermsthatareunlquetOaparticularlanguagecanalsobeaccommodatedintheTopic
Class.Thatis,thetoplCmapOfeachlanguageisnotlimitedtoone-t0-Onetranslationofthe
basictopicmap.Forexample,theconceptofearthquake-disaster(震災)thatisuniqueto
J叩aneSeCanbeassociatedwithoneortwoconceptsinotherlanguages.
InadditiontothemultilingualtoplCClass,themultilingualtranslationcanbeonthe
OccurrenceandAssociation.ForexamplethetypeofdocumentinOccurrence(journalpaper,
map,etc)canbetranslatedintomultiplelanguages:
Map 地図Peta
Thelanguageoftheorlglnaldocumentsthatneedtobeaccommodatedcanbefairly
complex.Thisisimportantfわrgreyliteraturesuchasone-pagedocumentsinlocallanguagesthat
areusedduringdisasters.Manycountrieshavelocallanguagesinadditiontothenational
language.China,India,andIndonesiahavetensorhundredsoflocallanguages.Thiscanbe
implementedusingOccurrenceinTopicMap.ThelanguageintheOccurrenceidentifieswhat
languageistheorlglnaldocument.Languageintheoccurrencecanbeimplementedasa
hierarchysimilartotheTopicClass.
5.CollaborativeApproachtotheDevelopmentorKnowledgeM ap
Thedevelopmentofknowledgemapfordisastermanagementcollectionswillbea
collaborativeem)rtbetweenthedevelopmentteam(librarian,informationscientist)and
prospectiveusers,especiallydisastermanagementexpertsfromvariousdisciplines.The
followlnglSalistofprinciplesandactivitiesinthedevelopmentofknowledgemap:
・ Noreinventlngthewheel.Thereisabodyofknowledgefrompreviouseffortsthatcanbe
usedinthedevelopmentofknowledgemapsuchastaxonomies,metadata,vocabularies,
andontology.AmongthefirststepsinthedevelopmenteffTortistocollectrelevant
previousworksinknowledgemanagementofdisastercollectionthatcanbeusedasthe
basisfわrknowledgemap.
― ��� ―
・ Decidewhoaretheusersofthedisastercollectionandtheusers.Weneedtoidentifywho
aretheprlmaryusersthatinclude:researchers,professionals,andreliefagencies.Thisis
importantsincetheknowledgemapwillbeadigestoftheusers'collectiveknowledgein
disastermanagement.
・ Conductrequirementanalysisontheusers(researchers,professionals,reliefagencies,...)
・ DeveloptopICmapbasedontheusers'inputs:Topic,Association,andOccurrence.
・ EvaluateandtestthetoplCmap
・ Linkingthetopicmapwithrealcollections(suchasRobeUniversity'sEDMC)
Thedevelopmentofmultilingualknowledgemapandthedevelopmentofoccurrencelink
betweentheknowledgemapandtherealdocumentsinthecollectionrequlreSlgnificant
resources.WecanadoptcollaborativemethodsthathavebeensuccessfullyusedintheInternet
environment,suchasthephenomenalsuccessofWikipediaasanactiveandexpanding
dictionary.Thismethodcanbeusedbothinthedevelopmentofmultilingualknowledgemapand
theuseofknowledgemaplnrealdisastercollections.ThefbllowlnglSalistorbriefstepsthat
canbeused:
・ Developabasetopicmapinonelanguage(fTorexample,English)
・TranslateandexpandthetermstootherlanguagesuslngaWikipediacollaborative
method.
・AnnotationandoccurrencelinksaredoneusingWikipediaapproach(usingvolunteers
fromvariouscountriesandorganizationstodotheannotation).
6.1mtegrationorKnowledgeMapwithKobeUniversity'sEDMC
Asstatedinprevioussections,havingaslngleaslngleunifledrepositoryfordisaster
managementcollectionsisnotrealisticsincethecollectionsareownedandmanagedbyvarious
organizationsaroundtheworld.Amoreoptlmalapproachwouldbeaunifiedadvancedmetadata
intheformofknowledgemapthatisimplementedinafTederatedenvironment.Theproposed
knowledgemapisdesignedtosupportfTederatedinfわrmationretrievalsystemsusedbyvarious
― ��� ―
institutionsinmanycountries.Agoodexampleonhowtoimplementitwouldbetointegrateit
withRobeUniversity'sEarthquakeDisasterMaterialsCollection(EDMC).
6.1RobeUniversity'SEDMC
KobeUniversltylibraryhascollected48,060documentsrelatedtodisastermanagement,
especiallythetopicofHanshin-Awajiearthquake.AsofJune,2008,4,688(4462titles)have
beendigitizedandincludedintheGreatHanshin-AwajiEarthquakeDisasterMaterials
Collection(http://www.lib.kobe-U.ac.jp/eqb/).Non-Japanesematerialsaccountfor1043
documents(652titles).Thefわrmatincludesbooks,journals(mostlyperiodicalssuchas
magazines),newspapers,pamphlets,Piers,photos,maps,movingimages,audio,computer
resources,dissertations,andotherinternetresources(SeeTable1f♭rthedetailbreakdownofthe
totaldocumentsandthenumberofdocumentsthathavebeendigitizedfわreachfわrmat
categories).
Typesor Notdigitized Digitized Total %digitized
Collection (mumberordocuments) (numberordocuments)
1 Book 6,331 429 6,760 6.3
2 Jouma1 12,209 17 12,226 0.1
3 Newspaper 12,954 214 13,168 1.6
4 Pamphlet 5,349 938 6,287 14.9
5 Onepagedocument 5,876 3,053 8,929 34.2
6 Picture 87 21 108 19.4
7 Map 132 4 136 2.98 Movie 252 ll 263 4.2
9 Soundrecord 84 2 86 2.3
10 Computerresource 81 2 83 2.4
ll Ph.Dthesis 0 0 0
12 WWW 7 7 14 50.0
Table1.NumberordocumentsinRobeUniverslty'scollectionandthepercentageofdigitizedversionthathasbeenincludedintheonlineEDMCsystem.
― ��� ―
Table1showsthatonlysmallpercentageofthedisastercollectionhasbeendigitized.
Thepercentageofdigitizeddocumentisslgnificantonlyfわrone-pagedocument,pamphlet,and
plCtureS.Oneseriousissueondigitizationthatneedstobeaddressediscopyright.Thecopyright
lawdoesnotallowRobeUniversltytOdigitizethematerialsandmakeitavailablefわrpublic.
ThisissueneedstobefurtherinvestlgatedastowhatthelimitofthecopyrightlawinJapanand
intheworldingeneral.Thereareanumberofpotentialsolutionsthatneedtobeexplored,
including:
・ MakethedigitizedfわrmavailableonlyfromKobeUniversltycomputers,Similartothe
electronicJOurnalsubscript10nSfrommajorPublisher.Itisnotclearifthissolutionis
allowedundercopyrightlaw
・ Usethedigitizefわrmonlyfわrdevelopingadvancedmetadata(suchasknowledgemap).
Inotherwords,thedigitaldocumentisonlyusedinternallyfわraidinguserretrieval,but
notavailabletotheusers.Thisismorelikelyallowablesincethedigitalfわrmcannotbe
accessedbyusers,althoughtheusefulnessofthisapproachisstillnotclear.
400350300250200150100500 l007
■
■
Figure4.Numberofcitationsof"Hanshin-Awaji"keyword,inGoogleScholar
ThenumberofdocumentsinthecollectionsisexpectedtocontinuouslylnCreaSedinthe
yearstocome.Approximately60documentsareaddedtothecollectioneverymonth.Using
GoogleScholartogaugethetrendofcitationsrelatedtoHanshin-Awaji,asearchusing
― ��0 ―
"Hanshin-Awaji''wasperfわrmedandtheresultsareshowninFigure4.TheresultisdiffTerent
thanpreviousstudyonMt.Helenacitation(Yeheta1.,2007)thatfわundthecitationpeakedinthe
firstthreeyears,followedbyrapiddeclineinthesubsequentyears.TheHanshin-Awajicitation
seemstobegolngStronga洗ertenyears,almostatthesamelevelasthefirstthreeyears.
AnewgenerationofEDMCretrievalsystemiscurrentlybeingdevelopedbyAriki
LaboratoryatKobeUniversity.Figure4illustratesthenewEDMCfull-textretrievalcapabilities.
Thefull-textindexisconstructedusingOCRapproachduetothefactthatthedigitized
documents,althoughinPDFfわrmat,areactuallylmageSnotreadablebyanyso氏ware.Thenew
systemusesfull-textindexthatwouldallowuserstoretrievedocumentsuslngkeywordsfrom
thedocumentcontents,1nSteadofbeinglimitedtothecurrentmetadatasuchastitleorauthor.
Thisnewapproachwillincreasetherecalloftheretrieval:itretrievesrelevantdocumentsthat
otherwisearenotretrievablewhentheretrievaltermisnotinthemetadata.However,as
discussedinsection4.2,full-textindexmayleadtohighnoiseandlowprecision.
Thefu11-textsystemthathasbeendevelopedbyArikiLaboratorycanbeveryusefulfor
thedevelopmentofadvancedmetadatasuchastheproposedknowledgemap.ThefuHtext
systemcanbeusedfわrautomatedmetadatageneration.Theproposedknowledgemapmetadata
canbepopulated(linkingthetopictodocument)usingacombinationofautomatedmethodand
manualcheckingtoimprovethequalityofthemap.
― ��� ―
Figure3.NewEDMCretrievalsystem(Sako皮Ariki,2008)
AnissuethatcameupduringdiscussionbetweenRUCSSandRobeUniversityLibrary
staffisthatsomeoftheJapanesekeywordswillbemeanlnglessiftranslatedintoEnglish
becausetheyareunlquetOJapanesesituation.Asdiscussedinsection4.5,thissituationcanbe
accommodatedintheproposedknowledgemapsincethetranslationm叩plngWillnotbesimple
one-tO-one.
6.2Recommendations
Theoverallgoalistodevelopaunirledadvancedmetadatafordisastermanagement
whileatthesametimeimprovlngEDMCsystematKobeUniversity.Theunifiedadvanced
metadatawillbeusedbycollaboratlnglnStitutionsaroundtheworldandKobe'SEDMCwillbe
thepioneerinadoptlngtheadvancedmetadatainanexistlnginfわrmationretrievalsystem.The
followlnglSalistofsuggestedstepstoachievetheoverallgoal:
・ DevelopknowledgemapfわrimprovingEDMC.ThedevelopmentofknowledgemapIS
inlinewiththeproposaltoJapaneseFoundationon"NewApproachtoDisaster
ManagementandRiskReduction".Iffunded,atleastpreliminaryworkonthis
developmentcanbedoneaspartoftheJapanFoundationproject.
・ Moresubstantialfundingwillbeneededtodevelopthisknowledgemap.Alarger
collaborativeproposalneedtobesubmittedtofundingagenciesbasedontheinitialwork
describedabove.
・ Addressimportantissuessuchascopyrightissuesfわrdigitization,whoshouldbedoing
digitization,etc.Sincedigitizationitselfisaslgnificantundertaking,perhapsitisbetterto
beconductedbyKobeLibraryInsteadofRUCSS.
・ Startcollaborationwithothercountries.AproposaltotheWorldBank'sSouth-to-South
programincludesadevelopmentofSumatra-Andamantsunamicollections.Sincethis
proposalwillacollaborativeeffTortsbetweensixcountries(Malaysia,Indonesia,Sri
Lanka,India,andThailand),itisanidealplacefわrcollaboration.
・ IntegratetheknowledgemapwithEDMCretrievalsystemanduseEDMCsystemto
evaluatethequalityoftheknowledgemap.Thisapproachwillimprovethequalityofthe
knowledgemapandtheretrievalperfわrmanceorEDMCsystem.
― ��� ―
References:
Britton,N.R.(1999).Whithertheemergencymanager?InternationalLTournalofMass
Emergencies&Disasters,17(2),223-235.
DiMaio,P.(2007)Anopenontologyf♭ropensourceemergencyresponsesystem.Availableat:
http://oDenSOurCe.mit.edu/papersrrOWARDSAN OPEN ONTOLOGY FOR ER.Ddf
Green,WGandMcGinnis,SR .,(2002)Thoughtsonthehigherordertaxonomyanddisaster.
NotesontheScienceofExtremeSituation,P叩erNo.7Availableat
http://facultystaff.richmond.edu/~wEreen仲APER7.pdf.AccessedJuly15,2008.
Ilalliday,Q,(2007)"DisasterManagement,OntologyandtheSemanticWeb",availableat
http://polaris.gseis.ucla.edu/qhallidaDisasterManagementFinal.html.
Joshi,H,Seker,R,Bayrak,C,Ramaswami,S.,andConnely,J.(2007)Ontologyfわrdisaster
mltlgationandplannlng.Proceedingsofthe2007summercomputersimulation
conference,SamDiego,Califomia.
Kreps,GA,Description,(1989)taxonomy,andexplanationindisasterresearch,Intemational
JoumalofMassEmergenciesandDisasters;7(3):277-80,Nov.1989.
Lehman,H.andWebster,∫.(2004a)DescribingGreyLiteratureAgain:ASuⅣeyofCollection
Policies.SixthInternationalConferenceonGreyLiterature:workongreylnProgress,6-
7December2004.D.∫.FaraceandJ.Frantzen,comp.Availableat:
http://ir.library.oregOnState.edu/dspacen)itstream/I957/15/1/Lehman-Webster.pdf
Lehman,H.andWebster,∫.(2004b)FinalReporttoOregonStateUniversityLibrariesandthe
NorthwestAlliancef♭rComputationalScience皮Engineerlng.OregonUniverslty
TechnicalReport.OSUEECSTechnicalReport04-40-04.Availableat:
http://ir.library.oregonstate.edu/dsDaCen)itstream/1957/25/l/TsunamiNAFR.doc
Mathews,B.S.(2004)Grayliterature:Resourcesfわrlocatingunpublishedresearch,C&RLNews,
March2004Vo1.65,No.3.
― ��� ―
Pepper,S.,171eTAOofTopicMaps;Pepper,Steve,Ontopia,April2002,
http://www.ontopia.net/topicmaps/materials/tao.htmI.
Ouarantelli.E.L.(1987)WhatShouldWeStudy?QuestionsandSuggestionsfわrResearchers
AbouttheConceptofDisasters,IntemationalJoumalofMassEmergenciesand
Disasters,5:7-32.
Rath,H.TopicMapsHandbook,availableat
http://www.emDOlis.com/downloads/empolisTopicMaDSWhitepaper20030206.pdf
Sotoodeh,M.Kruchten,P.(2008)AnOntologicalApproachtoConceptualModelingof
DisasterManagement,SystemsConfTerence.20082ndAnnualrEEE,MontrealQuebec,
7-10April2008.Pp:ト4
UscholdM.andGruningerM.(2004),Ontologiesandsemanticsfわrseamlessconnectivity.
JoumalofSIGMODRec.,Vol.33(4),pp.58-64.
Xing,W,Dikaiakos,DM,Yang,W.,andSphyris,A.,Buildingadistributeddigitallibraryfor
naturaldisastersmetadatawithgridservicesandRDF",Libra7yManagement,26:4/5,
2005,230-245.
Yeh,H.,Manson,C,andVeytser,V.,(2007).TsunamiGeologicDataRepository.Availableat: