NASA Technical Reports Server (NTRS) 20120010233: From Science to e-Science to Semantic e-Science: A Heliosphysics Case Study

0.29 MB·English

by  NASA Technical Reports Server (NTRS)

Computers&Geosciences](]]]])]]]–]]]
ContentslistsavailableatSciVerseScienceDirect
Computers & Geosciences
journal homepage: www.elsevier.com/locate/cageo
Review
From science to e-Science to Semantic e-Science: A Heliophysics case study
Thomas Narocka,b,n, Peter Foxc
aAdnetSystems,Inc,Rockville,MD,USA
bNASA/GoddardSpaceFlightCenter,HeliosphericPhysicsLaboratory,USA
cRensselaerPolytechnicInstitute,TetherlessWorldResearchConstellation,USA
a r t i c l e i n f o a b s t r a c t
Articlehistory: Thepastfewyearshavewitnessedunparalleledeffortstomakescientificdatawebaccessible.TheSemantic
Received30November2010 Web has proven invaluable in this effort; however, much of the literature is devoted to system design,
Receivedinrevisedform ontologycreation,andtrialsandtribulationsofcurrenttechnologies.Inordertofullydevelopthenascent
15November2011 fieldofSemantice-Sciencewemustalsoevaluatesystemsinreal-worldsettings.Wedescribeacasestudy
Accepted16November2011
withinthefieldofHeliophysicsandprovideacomparisonoftheevolutionarystagesofdatadiscovery,from
manualtosemanticallyenable.Wedescribethesocio-technicalimplicationsofmovingtowardautomated
Keywords: andintelligentdatadiscovery.Indoingso,wehighlighthowthisprocessenhanceswhatiscurrentlybeing
Semantice-Science donemanuallyinvariousscientificdisciplines.OurcasestudyillustratesthatSemantice-Scienceismore
SemanticWeb
than just semantic search. The integration of search with web services, relational databases, and other
Ontology
cyberinfrastructure is a central tenet of our case study and one that we believe has applicability as a
Provenance
generalizedresearchareawithinSemantice-Science.Thiscasestudyillustratesaspecificexampleofthe
benefits, and limitations, of semantically replicating data discovery. We show examples of significant
reductionsintimeandeffortenablebySemantice-Science;yet,wearguethata‘‘complete’’solutionrequires
integratingsemanticsearchwithotherresearchareassuchasdataprovenanceandwebservices.
&2011ElsevierLtd.Allrightsreserved.
Contents
1. Introduction........................................................................................................1
2. Setting:NASAHeliophysicsdataenvironment.............................................................................2
3. Casestudy .........................................................................................................3
3.1. Science......................................................................................................3
3.2. e-Science ....................................................................................................3
3.3. Semantice-Science.............................................................................................3
4. CurrentSemantice-Sciencelimitationsandinfrastructuralchallenges..........................................................4
4.1. Provenance...................................................................................................4
4.2. Ontologymappingandontologyintegration ........................................................................5
5. Summaryandconclusions.............................................................................................5
References .........................................................................................................6
1. Introduction aidsindatadiscoveryandintegrationandseveralimplementations
canbefoundintheliterature(Madinetal.,2007;Neumann,2005;
ManyorganizationsareleveragingtheInformationAge(Heyand Foxetal.,2007).Specifically,withinthefieldofheliophysicsthree
Trefethen, 2005) to move online to find new avenues of reaching SemanticWebenabledsearchandintegrationsystemsareavailable
their consumers. The past few years have witnessed unparalleled (McGuinnessetal.,2007a;Narocketal.,inpress).
efforts (Dalton, 2007) to make scientific data web accessible However, much of the literature is devoted to system design,
especiallyinthephysicalsciences.TheSemanticWeb(Berners-Lee ontologycreation,andtrialsandtribulationsofcurrenttechnologies
etal.,2001)hasproveninvaluableinthiseffort.TheSemanticWeb (e.g.Madinetal.,2007;McGuinnessetal.,2007a;Narocketal.,in
press). Moreover, Semantic e-Science is emerging as its own
discipline (Fox and Hendler, 2009a; McGuinness et al., 2009) in
nCorresponding author at: NASA/Goddard Space Flight Center, Code 672,
which intelligent applications automate scientific research tasks.
GreenbeltMD20771,USA.Tel.:þ13012861086.
E-mailaddress:[email protected](T.Narock). Theaforementionedimplementationanddeploymentstudiesserve
0098-3004/$-seefrontmatter&2011ElsevierLtd.Allrightsreserved.
doi:10.1016/j.cageo.2011.11.018
Pleasecitethisarticleas:Narock,T.,Fox,P., Fromsciencetoe-SciencetoSemantice-Science:AHeliophysicscasestudy.Computers&
Geosciences(2011),doi:10.1016/j.cageo.2011.11.018
2 T.Narock,P.Fox/Computers&Geosciences](]]]])]]]–]]]
asvaluablefirststeps;yet,weneedtocomplementthesetechnol- discovery methods. Following the recent trend of online data
ogy demonstrations in order to move Semantic e-Science main- discovery, and wanting to retain the uniqueness of each of these
stream. We need to investigate the socio-technical implications of sub-disciplines, NASA, in 2007, commissioned several federated
semanticsearchandtheeffectsonendusers.Technologyresearch informationsystems.Fiveinformationsystemswerefundedserving
and user behavior are not dichotomous, they are inseparable (Lee, thesolar,heliospheric,3magnetospheric,4Earth’sradiationbelts,and
2000).InordertofullydevelopSemantice-Sciencewemusttakea Earth’s upper atmosphere communities. Each of these systems is
design-scienceapproach(Hevneretal.,2004)inwhichweinvesti- responsible for providinguniform accessto their respective under-
gatenotonlyhowasystemworks,butalsowhyitworks.Evaluation lyingsourcesofheterogeneousanddistributeddata.Thesesystems
ofsystemsinreal-worldsettingsisinvaluabletothisprocess. arebroadlyknownasVirtualObservatories,aparadigm(Szalayand
ArecentNationalScienceFoundationreport(Cummingsetal., Gray, 2001) that began in astronomy and quickly spread to helio-
2008) concluded that ‘‘few studies integrate the distinct social, physics, oceanography, volcanology, and other diverse scientific
organizational,andinfrastructuredimensionsofdynamicdistrib- communities. Specifically, the Virtual Observatory paradigm unites
uted collaborations’’. The Cummings report (2008) also touches largequantitiesofdisparateandheterogeneousdatausuallyunder
onthephysicalsciencesasakeyapplicationareaforthistypeof one web-based portal. The underlying data remain heterogeneous
study.In regards tosemantictechnologies,thephysicalsciences and distributed, yet common metadata, access protocols, and
are amongst the early adopters of semantic technologies (Finin terminology provide transparent access to users. Within the NASA
andSachs,2004)andservetobenefitgreatlyfromit. Heliophysics domain the five Virtual Observatories are supported
In this work we describe a case study within the field of by numerous data analysis and visualization capabilities,5,6 that
Heliophysicsandprovideacomparisonoftheevolutionarystages enabletheprovisionofadiverseHeliophysicsdataenvironment.
of scientific research. Specifically, we start with a recent Helio- The systems within NASA’s Heliophysics Data Environment
physics study conducted under traditional means—manual implement search capabilities relevant to their domain. For
inspection of large quantities of data and personal interaction example, the Virtual Solar Observatory,7 dealing primarily with
with data producers and other domain experts. We label this solar images, focuses on optical search parameters such as
effort science as it describes the norm in heliophysics over the wavelengthandintensity.TheVirtualHeliosphericObservatory,8
pastfewdecades.ThedatadiscoveryportionofthisHeliophysics by contrast, deals primarily with in situ time series data. In a
study is then reproduced using multiple online search and similarmannertheremainingNASAVirtualObservatoriesimple-
retrieval systems. This effort is termed e-Science and describes ment search capabilities analogous to the types of data they
the recent trendto automate data discovery using webservices, contain.
databases,andotherInternettechnologies.Finally,wereproduce The diversity of scientific data and numerous methods by
thedatadiscoveryasecondtimeusinganewsemanticsearchand which to relate it makes for a challenging search problem.
retrievalsystem.WelabelthisscenarioSemantice-Science. Previous research (Merka et al., 2008a; King et al., 2008) into
Thesciencescenarioisusedasabaseline.Itprovidesuswith the applicability of traditional information retrieval techniques
an estimate of the status quo. It tells us the processes by which implemented in contemporary internet search engines (e.g.
theheliophysicistsconductedtheirresearchandtheresultsthat Google, Yahoo, etc.) has shown that relevance scoring, such as
they achieved. Comparing the Semantic e-Science results with PageRank(Brinand Page,1998),whilesuccessful inwebsearch,
scienceande-Scienceallowsustoseehowwellsemanticsearch fails in scientific search scenarios as there are few predefined
enhancesthescientificprocess.Inotherwords,weareprimarily connections between data. Any connections, if they exist, are
focusedonhowwellsemanticsearchenhancestraditionalscien- determinedduringtheresearchprocessbydomainexperts.
tific data discovery. We are interested in the socio-technical The recently emerged microformats9 are also impractical for
implications of moving toward automated intelligent discovery scientific search. Specifically, microformats offer a means to
andweseektounderstandhowwellthisprocessenhanceswhat augmentXHTMLwithwell-definedsemantics.Microformatshave
iscurrentlybeingdonemanuallyinvariousscientificdisciplines. seenextensiveusagelatelyasthenumberofwebsitesusingthem
The novelty of this research is that we begin to understand the hasbeenestimatedtobeinthehundredsofmillions.10However,
broaderimpactsofmovingSemantice-Sciencemainstreamandin theuseofmicroformatsimpliessearchoverwebpages,whichis
doing so we are able to come full-circle and report on the not how search is conducted within the sciences (Merka et al.,
relationshipbetweensemantictechnologiesandtheirefficacyin 2008a, King et al., 2008).The aforementionedlack of predefined
scientificresearch.Wealsoprovideempiricalevidenceinsupport connections between data and the fact that scientific data, and
ofseveralfoundingpropositionsofSemantice-Science. associatedmetadata,areoftennotstoredaswebpagesprecludes
Specifically we aim to address the following questions: How theuseofmicroformats.Inasimilarmanner,populartechniques
efficiently and effectively does semantic search allow us to suchsocialtaggingandcollaborativefiltersarealsonotrelevant
facilitate ingrained data discovery techniques in one physical formanyscientificsearchandretrievalsystems.Tagsandrecom-
sciencearea?Doessemanticsearcharriveatthesame(orbetter) mendations can be useful in scientific collaborations; however,
conclusionsasresearchersusingmanualdatadiscoverymeans?Is theyareprimarilyusefulforinitialdiscoveryandlatterevaluation
semanticsearchbyitselfsufficientfordatadiscoveryinscientific of fitness for use, and not the primary means of data retrieval.
research?Whataffectdoessemanticsearchhaveontheresearch Rather, users are supplying a set of constraints on the data and
process?Whatgeneralstatementsofapplicabilitycanbemadeto the system is identifying time periods and data sets in which
otherscienceareas? those constraints are met. Thus, solutions involving domain
2. Setting:NASAHeliophysicsdataenvironment
3TheregionofthesolarsystemaffectedbytheSun.
4TheregionofspaceencompassedbytheEarth’smagneticfield.
TheNASAHeliophysicsdataenvironment1,2isbrokenintoseveral 5http://hpde.gsfc.nasa.gov/hpde_data_access.html.
sub-disciplineseachwiththeirspecializedinstrumentationanddata 6http://spdf.gsfc.nasa.gov/.
7http://virtualsolar.org.
8http://vho.nasa.gov.
1http://dx.doi.org/10.1029/2009EO470001. 9http://microformats.org.
2http://hpde.gsfc.nasa.gov/. 10 http://microformats.org/blog/2007/06/21/microformatsorg-turns-2/.
Pleasecitethisarticleas:Narock,T.,Fox,P., Fromsciencetoe-SciencetoSemantice-Science:AHeliophysicscasestudy.Computers&
Geosciences(2011),doi:10.1016/j.cageo.2011.11.018
T.Narock,P.Fox/Computers&Geosciences](]]]])]]]–]]] 3
semantics and the use of logical inference are the ideal solution Table1
formostscientificsearchusecases. Spatio-temporalrestrictionsusedbySlavinetal.(2002)andMerkaetal.(2008b).
Combiningthediversityofdata,thediversityofsearchtypes, Therestrictionsforthequeriesusedinourcasestudy.
and the need to provide uniform online access we arrive at a
1. Measurementsoftypemagneticfieldoccurringintherange
scenarioripeforsemanticsearch.Semantictechnologies,suchas (cid:3)150o¼GSMX(Re)o¼15
the Web Ontology Language (OWL) (McGuinness and van 2. Measurementsoftypethermalplasmaoccurringintherange
Harmelen, 2004b) utilize the meaning of the data as well as (cid:3)15o¼GSMX(Re)o¼(cid:3)5
3. Measurementsoftypemagneticfieldoccurringintherange
relationships between constituent data. The ability of semantic
(cid:3)5o¼GSMX(Re)o¼(cid:3)2.5
search to enable subsumption, inheritance, and inferencing cap-
4. Magneticfieldandthermalplasmameasurementsoccurringinthe
abilities helps overcome data heterogeneity and terminology range20o¼GSMX(Re)o¼50
differences and allows encoding of previously unconnected con- 5. Restrictions1.,2.,and3.mustalsoberestrictedtotherange
cepts.Thelackofthesecapabilitiesinsyntacticsearchmakesits (cid:3)10o¼GSMY(Re)o¼10
6. Restriction4.mustalsoberestrictedtotherange
utilizationinscientificdatadiscoverylimited.
(cid:3)40o¼GSMY(Re)o¼40
Inparticular,twooftheNASAsystems(theVirtualHeliospheric
Observatory (VHO) and the Virtual Magnetospheric Observatory
(VMO)) now use formal ontologies to discover and integrate data
(Narocketal.,inpress).Moreover,thesetwosystemswerespecifi- 3.2. e-Science
callydesignedtosharetechnologyandinfrastructure(Merkaetal.,
2008a)inordertoreducecostsandshortendevelopmenttime.Asa While automated web technologies can reduce much of the
result,operationsofeachofthesesystemsarecompletelydrivenby aforementionedworkloadthequestionremainsastotheefficacy
the ontologies. In other words, there is one underlying software of such a system and technology’s ability to enhance the search
infrastructurewiththetwodomainsemanticscompletelycaptured process.Toexaminethisinacontrolledsettingwefirstreplicated
in independent formal ontologies. Simply pointing the software (Merka et al., 2008b) the Slavin (2002) study using the initial
to one of the ontologies changes the domain that is searched. versions of VHO and VMO. The initial systems used purely
McGuinnessetal.previouslydeveloped(2007a)asimilarframework, relational database technology converting web form input into
which may hint at a convergence of semantic deployment meth- Structured Query Language (SQL) queries. While these systems
odologies.However,thequestionsposedinSection1regardingthe hadmanyadvantages,reducingthesearchtimeto(cid:2)2h,theyalso
efficacyofthesesystemsstillremainlargelyunanswered. highlightedseveralsocio-technicallimitations.
The VHO and VMO integrate data within two distinct helio-
physics sub-domains. However, heliophysics research questions
do not always fall into one sub-domain or the other. Rather, in
3. Casestudy
today’s world of cross-disciplinary and system science, research
questions often encompass resources from multiple domains.
3.1. Science
Thus,researchersthinkintermsofcomplexinterrelatedsystems
whilemanagersareoften forcedto deploymultiplecomponent-
TheSlavinetal.(2002)studythatisthebasisofourcasestudy
based IT systems that are more manageable. Moreover, despite
represents research questions that are common within the field of
best efforts, users are often unaware of such divisions. In our
Heliophysics. Moreover, the Slavin (2002) study requires multiple
particular case, our spatio-temporal requirements span the sub-
typesofdatatobeavailableduringastringentsetofspatio-temporal
domains covered by VHO and VMO. A standardized messaging
conditions. Specifically, the Slavin study specifies a spatial config-
framework (Narock and King, 2008) allows results from one
urationoffourspacecraftwhilespecificinstrumentsonboardthese
system to be saved and used as a query in the other system.
spacecraftareoperatingandcollectingdata.Anyfourspacecraftare
However,usersmustfirstbeawareoftheexistenceandcapabil-
acceptable as long as they meet the spatio-temporal restrictions
ities of both systems. Subsequently, the researcher must artifi-
listed in Table 1. All restrictions must occur simultaneously. The
cially decompose their query into multiple questions of the
spatial restrictions are listed using the Geocentric Solar Magneto-
respectivesystems—aprocessthatisforeigntomostresearchers.
spheric (GSM) coordinate system—an Earth-centered system com-
Other limitations come behind the scenes of the information
moninHeliophysics.Suchaconfigurationofspacecraftisrareand
systems. In a relational search system all relationships must be
heliophysicistsdonotknowitsoccurrenceintervalsapriori.
stated explicitly. This simple fact has two far-reaching conse-
As a result, Slavin and colleagues (2002) selected the most
quences. First, it limits the discovery of new knowledge. Users
likely subset of spacecraft that might fit their needs. While this
mayfindrelationshipsunbeknownsttothem;however,theywill
selection was based upon extensive experience within the
not find a relationship that was not pre-defined by system
domain it nevertheless highlights the gap between data sources
maintainers.Semanticsearchalleviatesthislimitationandoffers
domain experts are familiar with and data sources that are
the potential for knowledge discovery—a capability essential to
regularly emerging and appearing online. In recent years this
scientific research. Second, the system maintainer who defines,
gap has been widening (Gray et al., 2005) to the point of being
approves,andcommitsthesepre-definedrelationshipsservesasa
acknowledgedinthepopularmedia(Hotz,2009).Withmore,and
bottlenecktothesystem.Thispersonslowsdowntheadditionof
increasinglyvoluminous,dataconstantlycomingonline,knowing
newdata,whichinturnslowsdownthescientificprocess.
whattosearchisbecomingintractable.Thisisincreasingtheneed
forsearchingeneralandsemanticsearchinparticular.
3.3. Semantice-Science
Slavin’s team manually investigated 16 months of an initial
data source to undercover 43 possible events. Gradually, these
The data search portion of the Slavin (2002) study was
events were whittled away as the remaining spatio-temporal
once again reproduced. By replacing the e-Science structured
restraintswereadded.Alltold,2useableeventswhereidentified
SQL queries with inferences over domain ontologies we sought
after (cid:2)100hofmanuallabor.11
to identify which areas of the scientific research process
are adequately addressed by semantic search and which parts
11Privatecommunicationwiththestudy’sfirstauthorDr.JamesSlavin. may still be lacking. For this process, we utilized the SPASE
Pleasecitethisarticleas:Narock,T.,Fox,P., Fromsciencetoe-SciencetoSemantice-Science:AHeliophysicscasestudy.Computers&
Geosciences(2011),doi:10.1016/j.cageo.2011.11.018
4 T.Narock,P.Fox/Computers&Geosciences](]]]])]]]–]]]
Such is certainly the case in semantic eScience. The addition of
basic Semantic Web technologies – ontologies, reasoner, and
softwaretoolkit(Jena15)–reduces thedatadiscovery process in
ourapplicationsfrom (cid:2)2hto (cid:2)15min.Suchareductionoccurs
because we no longer need to manually visit 2 sites and more
importantlywenolongerneedtoartificiallybreakourqueryinto
sub-domainspecific questions. In the Semantic e-Science imple-
mentation incoming queries are converted from a standard
XML representation (Narock and King, 2008) into SPARQL
(Prud’hommeauxandSeaborne,2008),asemanticquerylanguage
approvedasaW3C16Recommendation.Itisatthisstagethatthe
benefits of Semantic e-Science become apparent. Using our
domain ontology, which spans all of Heliophysics, we can infer
towhichsub-domaineachcomponentofthequeryrefers.Further
inferences can then be made as to which information system
withintheNASAdataenvironmentaddressesthissub-domain.At
present,thiscapabilityislimitedtotheVHOandVMOastheyare
the only systems utilizing semantics. Nevertheless, this limited
Fig. 1. Graphical depiction of system complexity within the e-Science and
Semantice-Sciencescenarios. capability illustrates an advantage over e-Science—namely the
artificialdecompositionofqueries.
ontology12 with its VHO13 and VMO14 instantiations. These
ontologies are OWL encodings (Narock et al., in press) of the 4. CurrentSemantice-Sciencelimitationsandinfrastructural
XML-based schema developed by the Space Physics Archive challenges
Search and Extract (SPASE) consortium (Harvey et al., 2008).
Specifically, these ontologies have Description Logic (Baader While basic Semantic Web technologies reduced the search
etal.,2003)expressivityALCOIN(D),35classes,20objectproper- time by orders of magnitude they are not sufficient to offer a
ties,45datatypeproperties,andontheorderof20,000instances. ‘‘complete’’answer.Thatis,theSemantice-Scienceapproachdid
The aforementioned SPASE consortium, which consists of notretrievealloftheresultsfoundwithinthescienceapproach.
heliophysicsresearchers,softwaredevelopersanddataproviders, While the Semantic e-Science approach offered previously
was founded to aid in the integration of heliophysics data. The unknown events to ponder, the semantic search only returned
intentofSPASEistocreateametadataschemathatwouldleadto oneofthetwooriginalSlavinevents.
standardized descriptions of NASA Heliophysics data resources. More importantly the results lacked provenance information
WithinthiscasestudytheOn-To-Knowledgemethodology(Staab regardingcertainaspectsofthequery.
et al., 2001) was used (Narock et al., in press) to continuously While the semantic search correctly identified data sets of
convertreleasesoftheSPASEmetadatamodelintoOWL.Itshould interest, calls to services that search the underlying numerical
benotedthatknowledgeacquisitionandknowledgerepresenta- valuesfailedtoturnuponeofthetwooriginalevents.Thiswas
tion have historically been challenging and time consuming due to the resolution of data that the service searched over and
efforts.However,On-To-Knowledge(Staabetal.,2001)andother wasnotafaultofanysemanticinferencing.However,thishigh-
emerging methodologies (e.g. Benedict et al., 2007; Fox and lights the importance of coupling semantic search to related
McGuinness, submitted for publication) are beginning to lead to researchinwebservicesandSemanticWebservices.Inorderto
quick deployment, rapid community buy-in, and reduced effort be completely successful these technologies need to be compo-
ontheend-userspart. nents in frameworks rather than independent streams of
We utilized the final versions of the VHO and VMO search research. This observation is consistent with previous research
systems that now utilize formal ontologies (Narock et al., in (Lim et al., 2010), which noted that e-Science lacks ‘‘a single
press). Fig. 1 illustrates the how the VHO and VMO search backbone that can support the entire spectrum of requirements
systemswereaugmentedbytheadditionofsemantics. thatareessentialtoundertakecross-disciplinaryscientificcolla-
EvidentfromFig.1isthatminimalchangesweremadetothe borations.’’Additionally,Abbott(2009)and(GoodmanandWong,
system architecture. Specifically, the system now consults the 2009) have commented that current and future scientific chal-
aforementioned ontologies prior to constructing SQL queries. lenges can only be addressed with a focus on systems thinking
Domainlogiciscapturedwithintheontologiesandcanbeeasily and generalized frameworks, respectively. Such groundwork has
manipulatedandinferencedwithusingSemanticWebtools.This been laid (Fox et al., 2009b) and is being advanced by many
isoppositeoftheinitialsystemdesigninwhichthisinformation groupsintheEarthandSpaceSciencescommunities.
hadtobehard-codedintothequeryalgorithms.Previousresearch
has shown (Narock et al., in press) that such a scenario reduces
4.1. Provenance
complexity, query execution time, and system maintenance. In
addition,wecanposeourcompletequerytothesystem,haveit
Also missing from the semantic search results is provenance
answertheportionsitiscapableofanswering,andinfertowhich
information. One definition of provenance is ‘‘the description of
othersystemtosendtheremainingportionsofthequery.
theoriginsofapieceofdataandtheprocessbywhichitarrivedin
In what has become a catch phrase of the Semantic Web,
a database’’ (Buneman et al., 2001, p. 316). Metadata does exist
Hendler (2007) mused, ‘‘A little Semantics Goes a Long Way’’.
foralldatasetsreturnedbythesystems,andthisinformationis
readilyavailabletotheuser.However,subtletiesintheexecution
12http://vho.nasa.gov/ontology/spase.owl.
13http://vho.nasa.gov/ontology/vho.owl. 15 http://jena.sourceforge.net/.
14http://vho.nasa.gov/ontology/vmo.owl. 16 http://w3c.org/.
Pleasecitethisarticleas:Narock,T.,Fox,P., Fromsciencetoe-SciencetoSemantice-Science:AHeliophysicscasestudy.Computers&
Geosciences(2011),doi:10.1016/j.cageo.2011.11.018
T.Narock,P.Fox/Computers&Geosciences](]]]])]]]–]]] 5
ofqueriescanextrapolatetothepointthatresultsareambiguous Observatory (VSTO) (McGuinness et al., 2007a, Fox et al.,
or even unusable. For example, our query required a spatial 2009b). Specifically, Slavin (2002) utilized an auxiliary data
restrictionthatwasexpressedbyakeyword—‘‘Near-EarthHelio- source in order to confirm initial assumptions. Such an effort
sphere’’. Within the domain semantics this term has specific could have been replicated via VSTO data; yet, that system
meaning and one can utilize various services to determine if a operateswithitsownindependentontologyandassociatedweb
spacecraftisinsideoroutsideofthisspatialregion.However,how services(Foxetal.,2007).Atpresentthereisnoformalmapping
thedataintheservicewasgeneratedisabsentfromtheresult,as betweenVSTOandVHO/VMOontologies.Thisisnotacriticismof
isalmostalwaystrueforcurrentweb-basedsearchengines;only eithersystem.Rather,itisanempiricalconfirmationofthebroad
limited provenance on the search result itself is often available. tendency (Cummings et al., 2008) to create one-off systems
The lack of formal identification that provenance metadata is without taking time or effort for harmonization and as a result
indeedrelatedto,andmustbereturnedwith,asearchresultisa the lack of interoperability (in this case, semantic interoperabil-
seriousomissioninthesearchforrelevantscientificdata.Absence ity)canhindersystem-levelscience.
of this information further leads to questions of accuracy as the Semantic e-Science needs to have a concerted effort to stan-
computationalmodels(e.g.Peredoetal.,1995;MerkaandSzabo, dardize,align,andmapontologies.Suchneedshaverecentlybeen
2004) that predict the spatial region have inherent biases, recognized with several National Science Foundation projects
assumptions, and errors known to the scientists who are most devoted to semantic interoperability, such as the SONET (Madin
likely to have initiated the search. This particular case is easily etal.,2007)project.AlsotheSemanticeScienceFramework(SESF;
remediedbysupplyingmoreinformationtotheresults,however, Fox et al., 2009b; McGuinness et al., 2010) project mentioned
asqueriesbecomemorecomplexthesesubtlepiecesofinforma- earlierhasamongitsgoalstoenableconfigurablesemanticdata
tionincreaseandtheuser’scompleteunderstandingoftheresults frameworks using modular approaches to ontologies and the
diminishes. bridging of disciplines as well as application level integration,
Provenance is an actively researched topic (Pinheiro da Silva also using ontologies to encode meaning and relations for such
et al., 2004) that is now finding its place on the Semantic Web applications (Rozell et al., 2010). Further, SESF advances the key
(McGuinness et al., 2007b), and specifically within semantic notionofsemanticallyawareapplication-leveltoolsasanessen-
eScience—known as knowledge provenance (Fox et al., 2008a, tialpartofthesemanticecosystem,integratingnotonlydataand
2008b;Zedniketal.,2009;Zedniketal.,inpress),andweseethis informationsourcesbutprovenanceaswell.
asakeycomponenttoprovidinguserswitha‘‘complete’’answer
to their queries. Complete in the sense that essential explana-
tions, verificationsand justificationsare given to satisfy a scien-
5. Summaryandconclusions
tist’squestionsaboutanyreturnedresult.
In the area of knowledge provenance, the encoding of prove-
Admittedly, one case study can only begin to address the
nance interlinguas such as the Proof Markup Language (PML;
questionssurroundingtheemergenceofSemantice-Science.Yet,
Pinheiro da Silva et al., 2004; McGuinness et al., 2007b) and the
our case study has shed light on emerging issues and provides
Open Provenance Model (Moreau et al., 2011) as ontologies
empiricalevidenceinsupportofsomeofthefoundingtenets(Fox
(Michaelis and McGuinness, 2010) brings new expressive power
and Hendler, 2009a) of Semantic e-Science. We now return to
to the important but previously disconnected provenance meta-
considerourinitialresearchquestions.
data.However,languagesforencodingarereallyonlyafoundation
for provenance. What really address a scientist’s curiosity are
(cid:4) Howefficientlyandeffectivelydoessemanticsearchallowus
application tools that work with the encoded provenance. At
tofacilitateingraineddatadiscoverytechniques?
present these tools include provenance specific search browse
(cid:4) Does semantic search arrive at the same (or better) conclu-
and visualization (McGuinness and Pinheiro da Silva, 2004a) as
sionsasresearchersusingmanualdatadiscoverymeans?
well as integrated applications such as provenance-aware smart
faceted search (Fox et al., in preparation). These tools work
primarily with PML but similar tools are geared to work with Semanticsearchandbasicsemantictechnologies(i.e.RDF/OWL,
OPM(e.g.theKARMAsystem;Caoetal.,2009).Manyofthesetools reasoners,andtoolkits)areprovidinggreatbenefittothescientific
however are still clumsy for users to use and much work to research process. As evidenced by this case study these basic
improvethemremains.Ofparticularsignificanceforconvergence technologies allow one to replicated previous studies in fractions
inthisareaofweb-basedprovenanceistheformationandactivity of the time. While this is but one of many possible metrics
ofaW3Cincubatorgroupforprovenance.17 (McGuinnessetal.,2007c)itisonethatbusyscientistsoftencare
This progress not-withstanding, ‘complete’ answers to user about most. Moreover, these technologies do not require the
queriesrequireamulti-domainknowledgebasethatisaninter- artificial decomposition of complex questions. Reasoning and
section and super-set of domain knowledge, provenance, and inferencing capabilities are providing great benefit to both the
oftenotherdataproductrelatedinformation(Zedniketal.,2009; user and the system maintainer. This study, as well as previous
Foxetal.,inpreparation). work(Narocketal.,inpress),hasshownevidenceofeasiersystem
maintenanceandeasierqueryexecutionwhentransitioningfrom
e-Science to Semantic e-Science although further quantitative
4.2. Ontologymappingandontologyintegration
studiesarerequiredtosubstantiatethisclaim.
While semantic search is a vital component of Semantic
The VHO and VMO share a common base ontology (Narock
e-Science it is not the only component. Interspersed withontol-
et al., in press) and as a result queries passed amongst the two
ogies are non-semantic technologies such as Web Services and
systemsareeasilyinterchangeableandexecutable.However,we
relationaldatabases.Howandwhenthesetechnologiesinteractis
foundthisnottobethecasewithinthebroadercommunity.Our
justbeginningtobeunderstood.Forexample,provenanceinfor-
casestudy querycould havebenefittedfromother datasources,
mation is often thought of as the lineage of the data and some-
specifically ones available via the Virtual Solar–Terrestrial
thingtopresentattheendofexecution.Yetthebenefitsofusing
provenancetoenhancethesearchprocesscanbeseeninprevious
17http://www.w3.org/2005/Incubator/prov/charter. research (Fox et al., in preparation) as well as in evidence
Pleasecitethisarticleas:Narock,T.,Fox,P., Fromsciencetoe-SciencetoSemantice-Science:AHeliophysicscasestudy.Computers&
Geosciences(2011),doi:10.1016/j.cageo.2011.11.018
6 T.Narock,P.Fox/Computers&Geosciences](]]]])]]]–]]]
presentedearlierinthispaper.Similarscenariosshouldbesought we have shown, one can closely replicate manual discovery
infutureresearch. methods,yeta‘‘complete’’answerwillrequiretheentireSeman-
Semanticsearchreturnedmixedresultswithinourcasestudy. ticWebstackandthefullemergenceofSemantice-Science.
While it pointed us toward previously unconsidered data it also
failedtofindoneoftheeventsfromthescienceapproach.Again,
this was not the result of the search processes itself, but rather References
the ongoing challenges of connecting semantic search to other
cyberinfrastructurecapabilities. Abbott,M.R.,2009.Anewpathforscience?Inthefourthparadigm:data-intensive
scientificdiscovery.MicrosoftResearch,111–116.
(cid:4) Issemanticsearchbyitselfsufficientfordatadiscoveryin Baader,F.,Calvanese,D.,McGuinness,D.L.,Nardi,D.,Patel-Schneider,P.F.,2003.
The Description Logic Handbook: Theory, Implementation, Applications.
scientificresearch? CambridgeUniversityPress,Cambridge,UK.
Benedict,J.L.,McGuinness,D.L.,Fox,P.,2007.ASemanticWeb-basedmethodology
Thiscasestudyindicatesthatsemanticsearchasimplemented forbuildingconceptualmodelsofscientificinformation.AmericanGeophysi-
calUnion(FallMeeting(AGU2007)(EosTransactionAGU88(52),FallMeeting
in our case study is insufficient for ‘‘complete’’ data discovery.
Supplement,AbstractIN53A-0950)).
This conclusion is reached for a number of reasons. First, the Berners-Lee, T., Hendler, J., Lassila, O., 2001. ‘‘The Semantic Web’’. Scientific
massivequantitiesofscientificdata(Grayetal.,2005)thatapply American(May17).
Berners-Lee, T., 2006. Artificial Intelligence and the Semantic Web, AAAI 2006,
toHeliophysics,makeitunlikelythatall,possiblynotevenmost, /http://www.w3.org/2006/Talks/0718-aaai-tbl/Overview.htmlS#(14).
datavalueswillresideasontologyinstances;letaloneasLinked Bizer,C.,Heath,T.,Berners-Lee,T.,2009.Linkeddata—thestorysofar.Interna-
Data(Bizeretal.,2009).Thus,semanticsearchisonlyasgoodas tionalJournalonSemanticWebInformationSystems5(3),1–22.doi:10.4018/
jswis.2009081901.
the cyberinfrastructure services to which it connects. This was
Brin, S., Page, L., 1998. The anatomy of a large-scale hypertextual Web search
evidenced in our study where the ontology returned the correct engineSource.ComputerNetworksandISDNSystems30(1–7)(April).
data set, however, data services were unable to retrieve all Buneman,P.,Khanna,S.,Tan,W.-C.,2001.WhyandWhere:ACharacterizationof
DataProvenance,DatabaseTheory,LectureNotesinComputerScienceBook
relevant time periods. Second, provenance plays an important
Series,vol.1973.Springer,Berlin/Heidelberg.
role in semantic search. Without the complete Semantic Web Cao,B.,Plale,B.,GirishS.,Robertson,E.,Simmhan,Y.,2009.Provenanceinforma-
stack (Berners-Lee, 2006, i.e. Provenance, Rules, and Trust) the tion model of karma, version 3. In: Proceedings of the IEEE 2009 Third
capabilitiesofsemanticsearchappeartobeartificiallylimited. InternationalWorkshoponScientificWorkflows(SWF’09).
Cummings,J.,Finholt,T.,Foster,I.,Kesselman,C.,Lawrence,K.,2008.BeyondBeing
There:ABlueprintforAdvancingtheDesign,Development,andEvaluationof
(cid:4) What affect does semantic search have on the research Virtual Observatories, Final Report From Workshops on Building Effective
process? VirtualOrganizations,NationalScienceFoundation,May,2008.Availableat:
/http://www.ci.uchicago.edu/events/VirtOrg2008S.
Dalton,R.,2007.Geophysicistscombineforces.Nature447(7148),1037.
Our case study has shown quantitative benefits in query Finin, T., Sachs, J., 2004. Will the Semantic Web change science? Science Next
execution time. Similar results were found in a previous study Wave.AmericanAssociationfortheAdvancementofScience(September15,
2004).
(Szabo et al., 2009) with researchers conducting studies in a
Fox, P., Zednik, S., West, P. Semantic provenance in support of explaining,
fractionofthetimepreviouslyrequired.Webelievethatthiswill justifyingandverifyingsciencedataontheweb.JournalofWebSemantics,
slowly,butsteadily,acceleratethepaceofheliophysicsresearch. inpreparation.
Fox, P., Hendler, J., 2009a. Semantic e Science: encoding meaning in next-
Time, money, and resources saved on one research problem can
generationdigitallyenhancedscience,inthefourthparadigm:data-intensive
bedevotedtootherproblems.Thecumulativeeffectisaccelera- scientificdiscovery.MicrosoftResearch,147–153.
tioninthecreationofnewknowledge. Fox,P.,McGuinness,D.L.,Cinquini,L.,West,P.,Garcia,J.,Benedict,J.,Middleton,D.,
2009b. Ontology-supported scientific data frameworks: the virtual solar-
terrestrial observatory experience. Computers and Geosciences 35 (4),
(cid:4) Whatgeneralstatementsofapplicabilitycanbemadetoother 724–738.
scienceareas? Fox,P.,McGuinness,D.L.,PinheirodaSilva,P.,Zednik,S.,Garcia,J.,DingL.,DelRio,
N., Chang, C., 2008a. Semantic provenance for image data processing. In:
ProceedingsofGeoinformaitcs2008DatatoKnowledge,June.
The benefits of using provenance to enhance the search Fox,P.,McGuinness,D.L.,PinheirodaSilva,P.,2008b.Knowledgeprovenancein
processhavebeenevidencedwithinthisstudyaswellasinother virtualobservatories:applicationtoimagedatapipelines.In:Proceedingsof
domains (Fox et al., in preparation). This tenet extends the theInternationalSemanticWebConference.
Fox,P.,Cinquini,L.,McGuinness,D.,West,P.,Garcia,J.,Benedict,J.L.,Zednik,S.,
traditionaldatalineageviewofprovenanceandappearstohave
2007.SemanticWebServicesforinterdisciplinaryscientificdataqueryand
generalapplicability.Moreover,Semantice-Scienceismorethan retrieval. In: Proceedings of the AAAI Workshop on Semantic eScience.
justsemanticsearch.Theintegrationofsearchwithwebservices, doi:10.11.144.7441.
Fox, P. McGuinness, D.L., 2011b. An Open-world iterative methodology for the
relational databases, and other cyberinfrastructure, has applic-
developmentofsemantically-enabledapplications.ComputersandGeosciences,
abilityasageneralizedresearchareawithinSemantice-Science. submittedforpublication.
Massivedatavolumes(Grayetal.,2005)prohibit,oratleastdeem Gray, J., Liu, D., Nieto-Santisteban, M., Szalay, A., DeWitt, D., Heber, G., 2005.
ScientificDataManagementintheComingDecade,ACMSIGMODRecord,vol.
it very unlikely, that all information will reside as ontology
34,Issue4December2005,(pp.34–41).
instances or Linked Data. Thus, specific and targeted research Goodman,A.A.,Wong,C.G.,2009.Bringingthenightskycloser:discoveriesinthe
intocomponentandframework-basedsystemsisvitallyneeded. data deluge, in the fourth paradigm: data-intensive scientific discovery.
MicrosoftResearch,39–44.
ThefutureofSemantice-Scienceisverypromising.However,
Harvey,C.,Gangloff,M.,King,T.,Perry,C.,Roberts,D.,Thieman,J.,2008.Virtual
there are numerous challenges that still need to be overcome. observatoriesforspaceandsolarphysicsresearch.EarthScienceInformatics
Semanticsearchispresentlyoneofthemaindrivers,however,it 1(1),5–13.
cannotexistinisolation.Weneedtomovebeyondclosedsystems Hendler,J., 2007.Thedarksideof theSemantic Web.IEEE IntelligentSystems
(January/Febuary).
tobroadersemanticaccess.Dataaccess,integration,provenance, Hevner, A.R., March, S.T., Park, J., Ram, S., 2004. Design science in information
andprocessingservicesallneedsemanticrepresentation(Foxand systemsresearch.MISQuarterly28(1),75–105(March).
Hendler, 2009a). This case study has shown the benefits, and Hey, T., Trefethen, A.E., 2005. Cyberinfrastructure for e-Science. Science 308
(5723),817–821(May).
limitations,ofsemanticallyreplicatingdatadiscoveryinonecase
Hotz,R.,2009.Adatadelugeswampssciencehistorians.TheWallStreetJournal
studyinthephysicalsciences.Ithasattemptedtoprovideamuch (August 28, 2009) available at:/http://online.wsj.com/article/
needed extension to the current technological discussion. SB125139942345664387.htmlS.
King,T.,Narock,T.,Walker,R.,Merka,J.,Joy,S.,2008.Abravenew(virtual)world:
Namely, we have investigated the socio-technical implications
distributedsearches,relevancescoringandfacets.EarthScienceInformatics
ofdeployingsemantictechnologiesinreal-worldapplications.As 1(1).doi:10.1007/s12145-008-0002-7(April).
Pleasecitethisarticleas:Narock,T.,Fox,P., Fromsciencetoe-SciencetoSemantice-Science:AHeliophysicscasestudy.Computers&
Geosciences(2011),doi:10.1016/j.cageo.2011.11.018
T.Narock,P.Fox/Computers&Geosciences](]]]])]]]–]]] 7
Lee,A.,2000.Systemsthinking,designscience,andparadigms:heedingthreelessons Moreau,L.,Clifford,B.,Freire,J.,Futrelle,J.,Gil,Y.,Groth,P.,Kwasnikowska,N.,
fromthepasttoresolvethreedilemmasinthepresenttodirectatrajectoryfor Miles,S.,Missier,P.,Myers,J.,Plale,B.,Simmhan,Y.,Stephan,E.,denBuss-che,
futureresearchintheinformationsystemsfield,keynoteaddress.In:Proceedings J.V., 2011. The open provenance model core specification (v1.1). Future
oftheEleventhInternationalConferenceonInformationManagement,Taiwan, GenerationComputingSystems27(6),743–756.
May. (available online at: /http://www.people.vcu.edu/(cid:2)aslee/ Narock, T., King, T., 2008. Developing a SPASE query language. Earth Science
ICIM-keynote-2000S. Informatics1(1).doi:10.1007/s12145-008-0007-2(April,2008).
Lim,H.B.,Iqbal,M.,Yao,Y.,Wang,W.,2010.Asmarte-Sciencecyberinfrastructure Narock, T., Yoon, V., Merka, J., Szabo, A., The Semantic Web in federated
forcross-disciplinaryscientificcollaborations.AnnalsofInformationSystems information systems: a space physics case study. Journal of Information
11,2010.doi:10.1007/978-1-4419-5908-9(Chapter3). TechnologyTheoryandApplication,inpress.
Madin,J.,Bowers,S.,Schildhauer,M.,Drivov,S.,Pennington,D.,Villa,F.,2007.An Neuman,A.,2005.Alifesciencesemanticweb:arewethereyet?ScienceSTKE
ontologyfordescribingandsynthesizingecologicalobservationdata.Ecology
2005(283),pe22.doi:10.1126/stke.2832005pe22.
Information2(3),279–296.
Peredo,M.,Slavin,J.A.,Mazur,E.,Curtis,S.A.,1995.Three-dimensionalposition
McGuinness,D.L.,Fox,P.A.,West,P.,Rozell,E.,Zednik,S.,Chang,C.,2010.Progress
and shape of the bow shock and their variation with Alfve´nic, sonic and
toward a Semantic eScience framework; building on advanced cyberinfras-
magnetosonicMachnumbersandinterplanetarymagneticfieldorientation.
tructure.EOSTransactions(IN22A-02).
McGuinness, D.L., Fox, P., Brodaric, B., Kendall, E., 2009. The emerging field of JournalofGeophysicalResearch100,7907–7916.
PinheirodaSilva,P.,McGuinness,D.L.,Fikes,R.E.,2004.Aproofmarkuplanguage
semantic scientific knowledge integration. IEEE Intelligent Systems 24 (1),
25–26. forSemanticWebservices.In:Bell,David,Bussler,Christoph,Yang,Jian(Eds.),
McGuinness, D.L., Fox, P., Cinquini, L., West, P., Benedict, J., Garcia, J., 2007a. TheSemanticWebandWebservices:InformationSystems.31(4–5),381–395.
CurrentandFutureUsesofOWLforScientificDataFrameworks:Successes Prud’hommeaux,E.,Seaborne,A.,2008.SPARQLQueryLanguageforRDF,W3C
andLimitations,InOWLED. Recommendation, 15 January 2008, available at: /http://www.w3.org/TR/
McGuinness,D.L.,Ding.L.,PinheirodaSilva,P.,Chang,C.,2007b.PML2:amodular 2008/REC-rdf-sparql-query-20080115/S,lastaccessed30August2009.
explanationinterlingua.In:Proceedingsofthe2007WorkshoponExplana- Rozell, E., Fox, P., Maffei, A., Belliau, S., 2010. S2S an application integration
tion-awareComputing(ExaCt-2007).Vancouver,Canada,July22–23. frameworkfordistributedoceanographic repositories. EosTransaction AGU
McGuinness,D.L.,P.Fox,L.Cinquini,P.West,J.Garcia,J.L.Benedict,D.Middleton, (FallMeetingSupplement,AbstractIN23A-1349).
2007c. The virtual solar-terrestrial observatory: a deployed Semantic Web Slavin,J.A.,Fairfield,D.H.,Lepping,R.P.,Hesse,M.,Ieda,A.,Tanskanen,E.,Østgaard,
application case study for scientific research. In the proceedings of the N., Mukai, T., Nagai, T., Singer, H.J., Sutcliffe, P.R., 2002. Simultaneous
19th Conference on Innovative Applications of Artificial Intelligence (IAAI). observationsofearthwardflowburstsandplasmoidejectionduringmagneto-
Vancouver,BC,Canada,July2007,pp.1730–1737andAImagazine,29,#1, spheric substorms. Journal of Geophysical Research 107 (A7). doi:10.1029/
pp.65–76. 2000JA003501.
McGuinness, D.L., Pinheiro da Silva, P., 2004a. Explaining answers from the Staab, S., Schnurr, H.P., Studer, R., Sure, Y., 2001. Knowledge processes and
Semantic Web: the inference web approach. Journal of Web Semantics ontologies.IEEEIntelligentSystems16(1),26–34.
1,397–413. Szabo, A., Merka, J., Narock, T.W., 2009. Multispacecraft observations of solar
McGuinness,D.L.,vanHarmelen,F.,2004b.OWLWebOntologyLanguageOver- wind gradients near L1: VHO case studies. American Geophysical Union
view, W3C Recommendation, 2004, available at: /http://www.w3.org/TR/
(FallMeeting2009,abstract#SH51B-1271).
owl-featuresS/.
Szalay, A., Gray, J., 2001. The World-Wide Telescope. Science 293, 2037–2038
Merka,J.,Narock,T.,Szabo,A.,2008a.NavigatingthroughSPASEtoheliospheric
(14September2001).
and magnetospheric data. Earth Science Informatics 1 (1). doi:10.1007/
Zednik,S.,Fox,P.,McGuinness,DeborahL.,PinheirodaSilva,P.Chang,C.,2009.
s12145-008-0004-5(April).
Semantic provenance for science data products: application to image data
Merka,J.,Szabo,A.,Narock,T.,Walker,R.,King,T.,Slavin,J.,Imber,S.,Karimabadi,
H., Faden, J., 2008b. Using the virtual heliospheric and magnetospheric processing. In: Juliana Freire, Paolo Missier, Satya Sanket Sahoo (Eds.),
observatoriesforgeospacestudies.EosTransactionsAGU89(53)(FallMeeting Proceedings of the First International Workshop on the Role of Semantic
Supplement,AbsractSA51B-05). WebinProvenanceManagement(SWPM2009),WashingtonDC,USA,October
Merka,J.,Szabo,A.,2004.Bowshock’sgeometryatthemagnetosphericflanks. 25,2009.CEURWorkshopProceedings526,CEUR-WS.org.
JournalofGeophysicalResearch109(A12224).doi:10.1029/2004JA010567. Zednik, S., Fox, P., McGuinness, D.L. System transparency, or how i learned to
Michaelis,J.,McGuinness,D.L.,2010.Towardsprovenanceandcollectiveknowl- worry about meaning and love provenance. In: Proceedings of the 3rd
edge for web applications. In: Proceedings of the Third International International Provenance and Annotation Workshop (IPAW2010), Troy, NY,
ProvenanceandAnnotationWorkshop(IPAW2010),pp.265–273. USA(June15–16,2010),LNCS,inpress.
Pleasecitethisarticleas:Narock,T.,Fox,P., Fromsciencetoe-SciencetoSemantice-Science:AHeliophysicscasestudy.Computers&
Geosciences(2011),doi:10.1016/j.cageo.2011.11.018