Escolar Documentos
Profissional Documentos
Cultura Documentos
ResearchObjects
FrommyExperiment
Contents
1ResearchObjects:TheBraindump
2FromWorkflowstoPacksonmyExperiment
3FromPackstoResearchObjects
4DefiningResearchObjects
4.1Replayablegobackandseewhathappened
4.2Repeatableruntheexperimentagain
4.3Reproduciblerunnewexperimenttoreproducetheresults
4.4Reusableuseaspartofnewexperiments
4.5Repurposeablereusethepiecesinanewexperiment
4.6Reliablerobustunderautomation
ResearchObjects:TheBraindump
DavidDeRoure,July2009
ThisisthemyExperimentstoryonResearchObjects.It'snottheonlytakeonthissubject!TheeLaboratory
groupshasbeenmeetingweekly,withrepresentativesofmyExperimentandmanyotherprojects,tobuildthe
biggerpicture.Thereareimportantmotivationscomingfromtheclinicaldomainandalsofromtheneedtohavea
commonobjecttoexchangebetweentheservicesthatmakeupeLaboratories.Wedon'taddressthesehere:thisis
ResearchObjectsaccordingtomyExperiment.
FromWorkflowstoPacksonmyExperiment
TheWeb2.0DesignPatterns(http://oreilly.com/web2/archive/whatisweb20.html)tellusthat"Dataisthenext
IntelInside",i.e.
Applicationsareincreasinglydatadriven.Therefore:Forcompetitiveadvantage,seektoowna
unique,hardtorecreatesourceofdata.
ThisremindsusthatthevaluepropositionofmanyWeb2sitesistheirexcellentsupportforonetypeofcontent.
Sowehavephotosonflickr(http://www.flickr.com),moviesonYouTube(http://www.youtube.com)andslideson
SlideShare(http://slideshare.net).And,whenwebeganmyExperiment,webecametheunique,hardtorecreate
sourceofscientificworkflowsspecificallyTaverna(http://taverna.org.uk)workflows.Weareproudtoholdthe
largestpubliccollectionofworkflowsavailable,andtohaveagrowingnumberofworkflowtypes,including
Microsoft'sTrident(http://www.microsoft.com/mscorp/tc/trident.mspx).
But,significantly,wealsorecognisedthataworkflowcanbeenrichedasasharableitembybundlingitwithsome
otherpieceswhichmakeupthe"experiment".Weobservethatresearchersdonotworkwithjustonecontenttype
andmoreoverthattheirdataisnotinjustoneplaceit'sdistributed,andsometimesquitemessytoo.Sowehave
http://wiki.myexperiment.org/index.php/Research_Objects 1/5
2/3/2017 ResearchObjectsmyExperiment
alsodevelopedsupportforpackscollectionsofitems,bothinsideandoutsidemyExperiment,whichcanbe
sharedasone.Ouruserinterfacemetaphorforpacksisfamiliarfromshoppingsiteslikeashoppingcartor
wishlist.
AnexcellentexampleofapackisPack55(http://www.myexperiment.org/packs/55),whichisPaulFisher'spack
forhistrypswork.Itcontainsworkflows,exampleinputandoutputdata,results,logs,PDFsofpapersandslides
asdepictedbelow.
Suchapackcapturesanexperiment,itcanbevalidated,it'sselfdescribedanditcanbeexportedwecouldsay
thatpacks"puttheexperimentintomyExperiment"!Thecollectionofpacksisgrowinginvarietyandnumber.For
examplethepackofallthepartsforapaper(http://www.myexperiment.org/packs/56)includesPaul'spack.There
arepacksofbenchmarkworkflows,packsofexampleworkflows,packsofpapersandslides.
FromPackstoResearchObjects
Aswehaveworkedwithourscientistsandresearcherswehavestudiedtheusecasesforpacksandwatchedhow
packsareactuallybeingused.Throughthiswehaverecognisedtheemergenceofanewformofdigitalobject
theResearchObject.Researchobjectsareanevolutionofpacksandprovidethesharable,reusabledigital
objectsthatenableresearchtoberecordedandreusedwhich,fundamentally,iswhatScienceandeResearch
involves.Infactitseemstousentirelylikelythat,inthefullnessoftime,objectssuchasthesewillreplace
academicpapersastheentitiesthatresearchersshare,becausetheyplugstraightintothetoolingofeResearch.It
isResearchObjectsratherthanpapersthatwillbecollectedinourrepositories.Aswellasaworkflowrepository,
myExperimenthasbecomeaprototypicalResearchObjectrepository.
ButwhatdotheseResearchObjectslooklike?Paulannotatedthepackwitharrowsandlabelsasshownbelow
thisisagoodexampleofwhatweneedinaResearchObject.
http://wiki.myexperiment.org/index.php/Research_Objects 2/5
2/3/2017 ResearchObjectsmyExperiment
AlsoDavidShottoninOxfordhasalsoprovidedanexcellentexampleofsemanticpublishing
(http://wiki.myexperiment.org/images/Shotton.png)whichcomesatthisfromtheotherendi.e.howdowe
augmentapaper.
Ourcurrentpackuserinterfacepermitsdescriptionsandcommentssothereisawayofexpressinginfreetextthe
relationshipofitemstoapack.Somedefactopracticesareemerging,suchashowpresentationsaredescribed.
Onepackcanbeincludedinanother.Butcurrentlywecan'treallyexpresstherelationshipsbetweenthingsin
packs,northerelationshipsbetweenpacksthemselves.Someofthoselabelswillbedomainspecificweare
implentingsupportforcontrolledvocabularies.Othersmaybemoregeneric,suchasbasicnotionsofprovenance.
Wehaveadatamodelwhichincludesworkflowsandpacksandwhichweareextendingtocapturethese
relationships.ObjectsfrommyExperimentcanbeobtainedinRDFfromrdf.myexperiment.org.Thedatamodel
usesamodularisedontologydesignedbyDavidNewmanwhichdrawsonDublinCore,FOAFandalsowithCarl
Lagoze'sOAIObjectReuseandExchange(http://www.openarchives.org/ore/)representation.We'reengagedwith
TimClark'sSWANSIOCwork(http://esw.w3.org/topic/HCLSIG/SWANSIOC),andalsowithLucMoreauand
hisOpenProvenanceModel(OPM)(http://openprovenance.org/)andYolandaGil'sproposedW3Cincubator
groupinprovenance.
Meanwhilewearealsogaining,fromourusers,asenseofwhatmakesagoodResearchObjectforexample,a
workflowcompletewithexampleinputandoutputdataprovidesameansofcheckingtheworkflowstilldoeswhat
itusedto(thankstoWernerMllerforhighlightingthisapproach).Onedayperhapswe'llhaveabarometertelling
youhowgoodyourResearchObjectis,likepasswordsecuritystrengthorfillingoutuserprofiles.
DefiningResearchObjects
TodefineaResearchObjectweneedtounderstanditscharacteristics.Weknowthattheyarefundamentallyabout
recordandreusewhichmeansrecordingresearchforanticipatedbutalsounanticipatedreuse,andthereinlines
thechallenge.
http://wiki.myexperiment.org/index.php/Research_Objects 3/5
2/3/2017 ResearchObjectsmyExperiment
We'veoccasionallyusedtheword"reproducible"inpresentations,andthereisaliteratureon"reproducible
research"(http://www.rrplanet.com/reproducibleresearchlibrum/viewtopic.php?f=30&t=77).WhenIwas
workingupmytalk(http://dl.getdropbox.com/u/1202407/Presentations/myExpESWC.ppt)forthepanelatESWC
2009(http://www.eswc2009.org/programmenu/panel)Itooktheprovocativepositionthatapaperisjustan
"archaichumanreadableformofaResearchObject",andthatinthefuturewewon'tsay"canIhaveacopyof
yourpaperplease"butrather"couldyousharethatresearchobjectwithmeplease?"TopresentthisIcreateda
slideofwordsbeginningwithR(infact,"Re.*")whichcharacterisebothfutureresearchandhenceaResearch
Object,andthishasevolvedthroughseveraltalks.
ThesearetheRsastheyarenow,onaWikipagesotheycancontinuetoevolveinpublic.Theyhavevariedin
numberbutsettleddownatsix:
Replayablegobackandseewhathappened
Experimentsareautomated.Theymighthappeninmillsecondsorinmonths.Eitherway,theabilitytoreplaythe
experiment,andtostudypartsofit,isessentialforhumanunderstandingofwhathappened.
Repeatableruntheexperimentagain
There'senoughinaResearchObjectfortheoriginalresearcherorotherstobeabletorepeattheexperiment,
perhapsyearslater,inordertoverifytheresultsorvalidatetheexperimentalenvironment.Andthescaleofdata
intensivesciencemeanslotsofrepetitionofprocessingtodealwiththedelugeofdataorindeedthedelugeof
methods.ResearchObjectsshouldhelpusscale.
Reproduciblerunnewexperimenttoreproducetheresults
Aseveryscientistknows,reproducibleisdifferenttorepeatable!Toreproduce(orreplicate)aresultisforsomeone
elsetostartwiththesamematerialsandmethodsandseeifapriorresultcanbeconfirmed.
Reusableuseaspartofnewexperiments
Oneexperimentmaycalluponanotheranexperimentmaybeusedinacontextotherthatthatinwhichitwas
originallyconceived.Byassemblingmethodsinthiswaywecanconductresearch,andaskresearchquestions,ata
higherlevel.Workflowsareagreatexample.
Repurposeablereusethepiecesinanewexperiment
Anexperimentwhichisablackboxisonlyreusaeableasablackbox.Byopeningthelidwefindparts,and
combinationsofparts,availableforreuse,andthewaytheyareassembledisacluetohowtheycanbereused.
Thismeanstheremustbeadequatedescriptioninsidethebox.Hencearesearchobjectisselfcontainedandself
describingitcontainsenoughmetadatatohavealltheabovecharacteristicsandforaeLabcomponentorservice
tomakesenseofit.(ComputerScientistshavean"Re"wordforthis"reflection",whichmeansyoucannotonly
runaResearchObjectlikeaprogrambutyoucanalsolookinsideitlikedata).
Reliablerobustunderautomation
Automationbringssystematicandunbiasedprocessing,andalso"unattendedexperiments"humanouttheloop.
Indataintensivescience,ResearchObjectspromotereliableexperiments,butalsotheymustbereliablefor
automatedrunning.
Retrievedfrom"http://wiki.myexperiment.org/index.php?title=Research_Objects&oldid=1521"
http://wiki.myexperiment.org/index.php/Research_Objects 4/5
2/3/2017 ResearchObjectsmyExperiment
Thispagewaslastmodifiedon1August2009,at06:31.
http://wiki.myexperiment.org/index.php/Research_Objects 5/5