Semantics for Web-Based Mathematical Education Systems
The
ActiveMathgroup:
Erica Melis, Jochen B ¨udenbender, Georgi Goguadze, Paul Libbrecht, Carsten Ullrich melis, jochen, george, paul, cullrich @activemath.org
Universit ¨at des Saarlandes, D-66123 Saarbr ¨ucken, Germany
ABSTRACT
Web-based user-adaptive learning environments suggest a
semanticknowledgerepresentationthatcanbereusedindif-
ferentcontexts. Moreover,iftheseeducationalsystemsem-
ployexternalservicesystemsforsupportorforexploratory
activities,thesemanticrepresentationisabasisfortheinter-
operability of the service systems and for machine-under-
standabledata. ActiveMath is such a learning environ-
mentformathematics. Weshowwhatitsannotatedseman-
tic knowledge representation, extended OMDoc, is like and
how it is used. We also discuss the current bottleneck of
authoring. Since mathematicians are mostlyfamiliar with
authoringL A
T
E
XratherthansemanticXML,ActiveMathof-
fersaL A
T
E
X2OMDoctool. AscomparedwiththedirectOMDoc
authoringwhichisnotyetvisuallysupportedthishaspros
andcons.
1. INTRODUCTION
Many educational systems and on-line documents have
been produced in recent years. Since the encoding of the
domainknowledgeforalearningenvironmentis averyex-
pensiveandtime-consumingtask,reusabilityoftheencoded
knowledge in dierent contexts and for dierent function-
alities is desirable. Therefore, the representation needs to
incorporate an ontology of the domain or, even better, a
uniqueandextensiblesemanticsofthedomainconceptsand
their variousrelationships. Similarly, inter-operability is a
prerequisiteformultipleservicesusedineducationsystems
thatcanaccessandworkwithcommonknowledgesources.
For Semantic Web applications, mathematics is a good
eld to experiment with because it is largely formalized
andhasaclearfundamentalsemanticsindependentofpre-
sentationalissues and because mathematics is a relatively
well-structured eld. For mathematics, anontology needs
to be enhanced by real semantics because mathematical
knowledgeisinherentlydierentfromitspresentation(e.g.,
itsprintedversion). Dierent presentations canmean the
same thing, e.g., 1
2
or 1/2. Conversely, the same presen-
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.
Semantic Web Workshop 2002 Hawai, USA Copyright by the authors.
tation canmeandierent thingsindierent contexts, e.g.,
( ab
p )=(
a
p )(
b
p
)isfalseinelementaryalgebrabuttrueinthe
theoryofquadraticresidues.
Now,ourlearning environmentActiveMath[8] isaSe-
manticWebapplicationformathematicslearning. Itsknowl-
edgerepresentationisseparatedfromitsfunctionalities. Its
knowledgerepresentationmeetstheaboverequirementsfor
mathematical contentrepresentations andthosefor educa-
tionalapplicationsthatincludetheabovementionedreusabil-
ityandinter-operabilityaswellastherepresentationofped-
agogicalinformation.
ActiveMath' knowledge representation is based on
OpenMath[3],ageneral,standardized,semanticXML-represen-
tation for mathematics. ActiveMath' functionalities re-
quire additionalinformationto beencodedintotheknowl-
edgerepresentation,e.g.,structuralinformationsuchasis-a-
denitionandpedagogicalinformationsuchasthediÆculty
ofanexercise.
ThisarticleshowshowSemanticWebissuessuchasmachine-
understandablerepresentation,reusability,extensibility,and
migration of otherrepresentations are tackled in Active-
Math. Itfocusesonthe knowledgerepresentation andits
currentauthoring. Itsummarizeswhichinformationisrep-
resented in the OMDoc-language which is an extension of
OpenMath. It describesandsubstantiatestheextensionswe
have added for theeducational andother purposesofAc-
tiveMath. It discusseshow content isauthoredpresently
inasituation,wheretoolsfor semanticrepresentationsare
emergingonlyand wherethehabitsofauthorsstill oppose
suchanencoding.
2. SEMANTIC REPRESENTATION
Althoughtoday'smostcommonrepresentationforknowl-
edgeofweb-basedsystemsisthesyntacticmarkuplanguage
HTML,forameaningfulreuseindierentcontextsandknowl-
edgesharingtheXMLrepresentationisessentialandtheRDF 1
frameworkwithdatarepresentingrelationsbetweenelements
canserveasabasisforbuildinganontology.
2.1 Semantics in Mathematical Knowledge
OMDoc has evolvedhistorically as a standard for mathe-
matical knowledgerepresentation,whichwedecidedto use
for oureducationalsystem. Becauseof thishistoryseveral
features havestilltobeadaptedtosemanticWebdevelop-
ments,e.g. RDF.However, abig advantageofusingOMDoc
isitstrulysemanticavour.
1
http://www.w3.org/RDF/
ematics? To ensurethe inter-operability of mathematical
systems,akeyword-annotationthatmightsuÆceforsimple
searchfunctionalitiesisnotsuÆcientanymore. Amachine-
readableinputformathematicalsystemssuchasComputer
Algebra Systems (CASs) and theorem provers requires a
mathematical semantics. That is, to provide a basis for
multiplesystems,theactualmathematicalobjects/formulas
have to berepresented. Candidates for the representation
ofmathematicalobjectsandconceptsaretheXML-languages
OpenMath 2
andcontent-MathML.
3
OpenMath isa(European)
standard for the semantic representation of mathematical
formula expressions. It semantically denesasetofmath-
ematicalsymbolsintheOpenMath content dictionariesand
canimportcontentdictionariesintoothers.
However,anextensionofOpenMathisneededbecause(1)
anontologybasedonOpenMathlacksmostrelationsexceptof
theory-inclusion, (2)theOpenMath contentdictionaries are
incomplete and a simple extension mechanism is needed,
(3) OpenMath has no means to structure the content of a
mathematicaldocumentbydividingitintoitslogicalunits
suchas\denition",\theorem",and\proof".
Forthesereasons,OpenMath hasbeenextendedtoOMDoc
[6]whichincludes structuremarkupfor mathematicalcon-
ceptssuchasdefinitionsandtheoremsandforotheritems
suchasexamples, exercises, and elaborative texts. It
alsoallowstodenenewsymbols. OMDocitemsmaycontain
metadata,formalelements,textualelements,andreferences.
Referencescanbeconceptidentiers,URLs,andadditional
codeelements. MetadatainOMDoc represent legalinforma-
tioncomplianttoDublinCoremetadata[11]andextradata
elementformetadataextensions. OMDocallowstorepresent
somemathematicaldependencies: morphismsbetweenthe-
ories,equivalenceofdenitions,proof-for. Implicitlyitcon-
tainsadependencyofsymbolsbytheoccurrenceofsymbols
inanothersymboldenition.
ForourapplicationtheOMDocmetadataandrelationsbe-
tweenelementsareinsuÆcient,because, ononehand,Ac-
tiveMath needs a pedagogical ontology. For its use in
learning environments, we have extended OMDoc. On the
otherhand,relations betweenelementssuchasmathemat-
icaldependency, corollary of a theorem, similar examples,
counterexamplefor aconcept, whichwe introducearealso
generalfortheeldratherthanduetotheeducationalappli-
cation,butarenotpresentinOMDoc. Someofourextensions
are not specic for tutorial applications, such as technical
requirements 4
andbibliographicalreferences.
2.2 Pedagogical Knowledge
Theextensionsdescribedinthefollowingaremotivatedby
the tutorialapplication and generallyapplicable for learn-
ingsystemsratherthanspecicforamathematicssystem.
5
Theyinclude,amongothers
pedagogicalpropertiessuchasdiÆculty. Theyarerel-
evantforActiveMath'user-adaptivitybecausethese
2
http://www.openmath.org/
3
http://www.w3.org/Math/
4
Forclient-serverapplications, theannotationshave toin-
cludethe technical requirementsto display or to invokea
material. Theyneedto beknown, inparticular,to ensure
thatmultimediamaterialisonlyoered,ifthecomputeron
thelearnersclientcanhandleit.
5
rentlearningsituation;
pedagogicallymotivatedrelationsamongthedierent
piecesofknowledgesuchasis-prerequisite. Thisinfor-
mationcanbeused,e.g.,topresentthelearnerprereq-
uisitesfor understandingaconcept,togeneratelinks,
and togenerate anadaptively chosenand structured
learningcontent.
Ontheonehand,metadatastandardsforlearningresources
6
containtoomanymetadatawhicharenotrelevantforour
purpose. Ontheotherhand,adaptivelypresentingcontent
requiressomemetadatawhicharenotyetspeciedinLOM.
Thisisnotsurprising,sinceIMS 7
willbesoonextendedby
EducationalModellingLanguage(EML) 8
metadata. There-
fore, the ActiveMath metadata extensions of the OMDoc
DTD include some metadata from LOM as well as some
others. Forexample,weextendthelistofpossiblevaluesof
thetypeofattributeofLOMmetadataelementrelationas
describedbelow. Moredetailed, the pedagogicalmetadata
denedintheActiveMath-DTDarefield,abstractness,
difficulty, learning-context whichbelongtothe OMDoc
inFigure1.
<definition id="def_order">
<metadata>
<Title xml:language="en">
Definition of the order of a group element
</Title>
<extradata>
<field use="mathematics"/>
<abstractness level="neutral"/>
<difficulty level="easy"/>
<learning-context use="univ_first_cycle"/>
<relation type="for">
<ref theory="Th1" name="order"/>
</relation>
<relation type="depends_on">
<ref theory="Th1" name="group"/>
</relation>
</extradata></metadata>
<CMP xml:language="en" verbosity="3">
... </CMP>
</definition>
Figure1: ExcerptfromanOMDocfor adenition
fielddescribestheeldtowhichthecontentoftheitem
belongs. ItenablesActiveMathtopresentitemsfrompar-
ticularelds(suchasstatistics,physics,oreconomy),ifthis
is requiredby apedagogical strategy. For instance,if stu-
dentsfromdierentgroupslearnstatistics,e.g.,technicians,
mathematicians,psychologystudents,theyobtaindierent
(motivating) examples and exercises from the appropriate
eld. The meaning of abstractness and difficulty is
self-explaining. They serve to adapt the document to the
learner'scognitivecapabilities andlearning progress. Cur-
rently,these metadatacanhaveoneof threedierent val-
ues. learning-contextspeciesforwhichlearningcontext
thematerialwasintendedoriginally. Thepossiblevaluesof
6
suchastheLearningObjectMetadata(LOM)standardized
byIEEE-LTSC(http://ltsc.ieee.org/)
7
http://www.imsglobal.org/
8
learning-context are those dened in the IMS-metadata
standards. This information is important incase material
fromdierentsourcesismergedforanewcourse. Further-
more,theOMDocinFigure1hasaverbosity-attribute. The
information aboutthe verbosityof the textualpartsallow
thegenerationofdierentkindsofdocumentsuchasslides
andmoreverbosescripts.
ThefollowingmetadatadenedintheActiveMath-DTD
characterizeexercisesmoreprecisely
the targeted mastery-level of the exercise. Its val-
uescanbeknowledge,comprehension,application,or
transfer
the taskofthelearner whichcanbecalculate,check,
explore,giveexample,model,orprove
levelofinteractivityandaveragelearningtime
thetechnicaltypeoftheexercise. TheActiveMath-
DTDallowsforthevaluesprovidegap,mupad,maple,
omega,multiple choice, andllincurrently.
ActiveMathemploysthesemetadatatouser-adaptivelyse-
lectexercisesfor adocumentandinthesuggestion mecha-
nismaccordingtoaparticularteachingstrategythattargets
aparticularmastery-levelandadaptivelysupportsskillac-
quisition.
Themetadatarelationisusedtorepresentseveralrela-
tionshipsbetweenOMDocitems. Thetypeofthisrelationis
speciedinthetype-attributewhichcanhavethefollowing
meanings:
the previous knowledge required to understand the
item. Forinstance,theelementinFigure1depends-on
establishesdependenciesonpreviousconcepts.
thesimilaritybetweentheitemandanotherone,e.g.,
for two examples, denitions, exercises etc. that are
similaras,e.g.,showninFigure2.
afor-relationshipswhichcanbeusedinordertochar-
acterizeanitemwithanadditionalfunctionality. For
instance,anitemthatis aproofforatheoremcould
aswellbeanexampleforamethodapplicationoran
examplecouldbeacounterexampleforaconcept.
acitation-relationshipreferringtoanadditionalbib-extra
elementwhichisdenedinActiveMath-DTDtoen-
able afullfeatured citationmechanism.
OnecanarguethatourXMLspecicationsare\heavyonat-
tributes". There are some pros and cons for this. Pros:
when using attributes one can x the default values for
them,whereastheforbodyofanelementwecanonlyspec-
ify the type of data that canbe placed inside (e.g. child
elements,PCDATAetc.). Cons: nodirectstandardscom-
pliance (translation needed). Also note that, when using
attributes,theneedtointerpretethelabelsisnotintroduc-
inganyadditionaleortforXMLdatamanipulationengines.
ThedevelopementofActiveMathmetadatawillbecon-
tinuedby: separatingthemetadataandthecontentdatabases;
IMS content packaging (as soon as EML is integrated in
<metadata>
<Title xml:language="en">
Definition of left cosets
</Title>
<extradata>
<relation type="for">
<ref theory="Th1" name="leftcosets"/>
</relation>
...
<relation type="similar">
<ref xref="def_rightcosets"/>
</relation>
</extradata></metadata>
<CMP xml:language="en">...</CMP>
</definition>
Figure2: Therelation toasimilar denition
3. SUPPORT FOR AUTHORING
AuthoringontologicalXMLisinvestigatedinseveralprojects,
e.g., the SemanticWeb project 9
and the Ontology Editor
Protege[9].
10
Authoringtoolsfor the describedtrulysemantic(math-
ematical) XML are still insuÆcientlydeveloped. Suchtools
willnotonlyhavetosupporttheauthorinemployingorex-
tendinga(mathematical)ontologybutalsoinauthoringor
choosingpedagogicalmetadata,inauthoringexerciseswith
external service systems, and support her in writing (ab-
stracted)mathematicalformulasthatlatercanbepresented
viastylesheets. Thisisaseriousbottleneckcurrently.
So,whataretherealisticalternativescurrently?First,use
astillpreliminarytool,QMath,brieydescribedinx3.1that
supports the authoring of mathematical formulas. QMath
provides atleastsome supportbutisnot suÆcientlycom-
fortable for the average author. Second, translate from a
syntactically oriented macro-based encoding and heuristi-
cally addsemanticsas describedinx3.2. Third, wait until
the open-source community including our group has pro-
ducedanicevisualtool(someofitexists already,e.g., our
visualeditorfortablesofcontent).
3.1 QMath
QMath is a tool and a migration format for producing
OMDoc documents. It was developed by Alberto Gonzales
Palomo and is currentlyused by ActiveMath authors to
writeOMDocs. AQMath documentis easytoproduce. This
ispartiallyduetothefactthatQMathsupportsunicodeand
theauthorisfreetowriteherformulasdirectlybyusinguni-
codesymbols. Look,for instanceatthefollowing example,
correspondingtotheL A
T
E
XsourcefromtheFigure4:
Definition:[<-df1]
:"The definition of cartesian product"
:for:[cartesian_product]
:depends_on:[def_pair]
...
$MN Df suchthat(pair(x,y),x 2M&y2N)$.
Figure3: Excerpt fromaQMathdocument
Theauthorcandeneacontextforadocument. A con-
text containsuser-denedshortcuts forOMDoc elements. It
alsocontainsthereferencesofsymbolsusedinthedocument
9
http://www.semanticweb.org/
10
theauthorusesasymbol,QMathsuggeststhesymboldecla-
rationprocedurewhichassignsameaningtothesymbolby
anexplicitreferenceto asymbolinacontentdictionaryor
viaapre-recordedcontext.
QMath supports metadata elements of OMDoc as well as
someadditional metadataofActiveMath. Thismetadata
supportcanbe extended. Assoonasanewelementis de-
claredinadocumentorinacontextQMathperformsatrans-
formationtoXMLmarkup.
11
3.2 L
ATEX2
OMDocForhandlingmacro-basedencodings,wefacetwodemands:
(1)the migration of existing mathematical learning docu-
mentsourcesencodedinpresentationallanguages,sayL A
T
E X,
and(2)thenewencodingbyauthors/mathsprofessorswho
areusedtowritingL A
T
E
Xandopposeauthoringatrulyse-
mantic representation. In the following, we describe our
eortsinbothdirections.
The migration ofan existing large L A
T
E
X documentwas
thegoal ofacase studywe conductedin2001. TheL A
T
E X
sources of the document [4] were not intended to be mi-
gratedtoasemanticrepresentationoriginally. Itunrestrict-
edlyusesauthor-specicmacros.AlthoughtheL A
T
E
Xsource
was alreadysplitinto 'slices'andprovidedsomedependen-
cies,thedocumentwasdesignedprettylinearlyratherthan
appropriateforahypertextpresentationanditwasdiÆcult
toprepareforareuse. Forinstance,itincludedtextsuchas
"Aswehaveseeninthepreviousexample...".
The logical structureof the in the existing L A
T
E X docu-
ment hadto be carefully redesignedin orderto obtain in-
dependentlyreusable items related via dependencies. For
instance, many basic pieces in the text still included an-
otherone. E.g., introductionsor elaborations containeda
denition.
Another problem was the use of not-represented abbre-
viations. Thetext contained elements like ... the correct
notation for this should be
@f
@x
(a) but we shall write only
@f
@x
sinceitis clearthat weare talkingaboutthederivative
inthepointa.Asemanticrepresentationwouldneedtore-
fertothesamemathematicalobjectthatmayhavedierent
annotations,e.g.,abbrevandadefault.
Thepurely syntacticuseofnotationis oneofthe major
problems. InasentencelikeBecarefullwithournotations!
In some cases (a;b) will mean an open interval and oth-
erwise just an ordered pair. (a;b) purely syntactical and
itssemantics iscontext-dependent. Anautomatictransla-
tionis almost impossibleor at least has ahighly context-
dependent heuristics. The presentation-oriented represen-
tation and mis-use of L A
T
E
X is another problem. This is
obviouswhentheL A
T
E
X encoding$\{x: x\in A$ und $x$
ist rational $\}$) of the formula fx: x 2 A und x ist
rationalg is analyzed and shows that a mathematical for-
mulaisscatteredintopiecesandcombinedagainbynatural
languagetext.
Apart the context-dependent and presentation-oriented
representation, presentationaland semantic information is
mixedespecially in the mathematical formulas and there-
foreheuristics for correctlyparsing all formulas cannotbe
11
FormoreinformationonQMathseehttp://www.matracas.
callyimpossible.
Ourexperiencesuggeststhatitisimpossibletotranslate
a L A
T
E
X source automatically that has been writtenwith-
outthegoalofasemanticrepresentationinmind. Thisnot
onlyrequirestoimplementmanydocument-specicheuris-
tics, it boilsdowntoabout50% manual translation. Even
worse,often theoriginal encoding doesnotallowa unique
translationtosemanticallyrepresentedformulasandmaybe
author'sworkagain.
A solutioncanonlyrely onaquasi-semanticmarkupin
L A
T
E
X that usesmacrosand environmentsto encodeinfor-
mationneededforsemanticknowledgerepresentationandis
extensible. Thishasbeenattemptedinanothercase study,
where weinstructed'conservative'authorsto writestrictly
denedL A
T
E
Xsourcesandprovidedatoolforanautomatic
conversiontoOMDocviaQMath.
WedenedL A
T
E
Xmacrosandenvironmentsfortherepre-
sentationofsemanticinformationandmeta-datatosupport
the automaticconversiontoOMDoc. Ifanelementhasnon-
emptychildren,itisencodedbyanenvironment,otherwise
a macro is used 12
. We speciedrestrictions { inparticu-
larforwritingformulasinthatL A
T
E
X. Themostimportant
restrictionsare
use thesymbolsalready denedinOpenMath content
dictionariesinordertoensurereusability,
specifytheinterpretationofsourceformulaswrittenin
L A
T
E
X,e.g. inxorprexnotation
use the pre-dened L A
T
E
X environments and macros
for dening OMDoc elements, i.e., structure elements
andmetadata.
ThefollowingisanexampleofusingspecialOMDoc-oriented
L A
T
E
Xenvironments:
\begin{definition}{df1}{carte sian _pro duct}
{The definition of cartesian product}
\depends-on{def_pair} ...
$M \times N \Df \suchthat{\pair{x}{y}}
{x \in M \and y \in N }$.
\end{definition}
Figure 4: Example ofwriting restricted L A
T
E X
create aseparate leto dene newsymbols and add
XSLpresentationalinformationtothedenedsymbols,
also dene own DTD extensionif necessary and XSL
presentationforit.
Iftheserequirementaremet,ourtoolautomatically con-
vertstherestrictedL A
T
E
XsourcesviaQMathtoOMDoc.
To summarize: as comparedwith a directauthoring of
OMDoc in QMath, authoring in a restricted and augmented
L A
T
E
Xismorefamiliartomathematiciansevenifnotstrictly
simpler. Althoughthedirectcontrol ofthelayoutofadoc-
ument by editing the generated PDF-document is very at-
tractivetoauthors,itkeepsthepresentationandloosesthe
representationandthusdestroysthesemanticandmetadata
informationneededfortheSemanticWebapplication.
12
For more details see http://www.activemath.org/~ilo/
4. USAGE IN
ActiveMathInsteadofaconclusion,wewanttosummarizewhatAc-
tiveMathisable todowiththeknowledgerepresentation
andwhatthefutureactivitieswillbeinthisdirection.
FortheActiveMathsystem,thereuseofcontentindif-
ferent contexts is particularly important because its user-
adaptivity implies thatthe samecontentcanbe presented
indierentwaysdependingontheuserandinthelearning
situation.
ActiveMath' user-adaptive functionalities such as the
presentationofthecontentandthedynamicsuggestiongen-
eration use the structureinformation and metadata anno-
tatingtheunitsofthecontent.
ActiveMath has the following components: a session
manager,coursegenerator,themathematicalknowledgebase,
a presentation planner, a user model, a pedagogical mod-
uleandexternalmathematicalsystems(ActiveMathinte-
gratesseveralservicesystemsforcalculation,proof,andex-
plorationsuchastheproofplannermega[7]andtheCom-
puterAlgebraSystemsMuPAD[10]andMaple). Here,the
usermodelis acomponentto store, readandupdate data
aboutthelearner'sprole. Itcontainshistoryoftheactions,
alistofpreferencesoftheuserandalistofcompetenceas-
sessments. The user's actions are analyzed by evaluators
thatcalculateupdatesoftheusermodel.
ThecoursegenerationinActiveMathisrealizedasfol-
lows: requests of the user are sent from the browser via
a web-server to the session manager. When the user has
chosenhergoalconceptsandscenario,thesessionmanager
sendsthisrequesttothecoursegenerator. Thecoursegen-
eratorcontactsthemathematicalknowledgebaseinorderto
calculatewhichmathematicalconceptsarerequiredforun-
derstandingthegoalconcepts. Thentheinformationabout
theuser'sknowledgeis requestedfrom theusermodeland
thecollected IDsofOMDoc itemsannotatedwiththeuser's
knowledgemastery-levelsareenteredasfactsintotheknowl-
edgebase of theexpertsystem. Thenthe rules are evalu-
ated and generate aninstructional list of XML items to be
presented. Here, the metadata such as diÆculty are not
onlyusedtoselectappropriateexercisesandexamplesfora
learnerbutalsofortheevaluationoftheuser'sactivities.
The XML content is transformedto aformat suitable for
presentationviaXSL.AnXSLstylesheetspeciesthepresen-
tationofourXMLdocuments,bydescribinghowaninstance
istransformedintoHTMLortoL A
T
E
XorFlash.
Thesemanticrepresentationisabasisformergingcontent
from dierent sources and presenting the merged content
consistently.
TheActiveMathknowledgerepresentation isproviding
twoontologies: themathematicalandthe educationalone.
Mathematicalontologyis alsousefulfor other mathappli-
cations. Themathematicalconcepts(elementsofontology)
aretheskeleton(macrolevel)ofadocument,andtheedu-
cationalonesprovidetheinformationforbuildingthemicro
levelstructure.
Alternative Usages
ActiveMath'supportforexploratoryandinteractivelearn-
ing will be improved. This includes the investigation of
elaborate search functions based on the the semantic and
partiallyformalrepresentation.
Anextstepisthemachine-understandabledescriptionof
many situations, includingthe automatedadviseto auser
forchoosinganappropriatesystemtoperformataskorfor
agent-basedcomputations(see[5]).
Semantically represented repositories will be useful not
justfor learningenvironmentsbutalsofor working mathe-
maticians. For instance, OMDoc-structured repositories can
improve the organization and searchability of mathemati-
cal knowledge. Today the digitallibraries,haveto face the
manuallycontrolledentryofauthorandclassicationinfor-
mationandoftenmakeuseofkeywordsauthoredbyreview-
ers. Todaythesearchcapabilitiesarelimitedtotextualand
keywordsearch.
13
WeunderstandourresearchaspartofthelargerEuropean
initiativefor web-based mathematical knowledgerepresen-
tationandmanagement. Itsrstworkshop[1] coveredtop-
icsrangingfrompublishingoflargecollectionsofelectronic
preprints to tools for managing mathematical documents.
Thereishopeforacriticalmassofcontentencodedinase-
manticXMLsincetheinitiativewillworkonthisaswellason
toolstomaintainandusethecontentdataandmetadata.
5. REFERENCES
[1] ElectronicProceedingsoftheFirst International
WorkshoponMathematicalKnowledgeManagement,
September2001.
[2] J.Budenbender,G.Goguadze, P.Libbrecht,E.Melis,
andC.Ullrich.MetadataforActiveMath.Seki
ReportSR-02-01,Fachbereich14Informatik,
UniversitatdesSaarlandes,2002.
[3] O.CaprottiandA.M. Cohen.Draftoftheopenmath
standard.OpenMathConsortium,1998,
http://www.nag.co.uk/projects /Open Math /omst d/.
[4] B.I.DahnandH.Wolters.AnalysisIndividuell.
Springer-Verlag,2000.
[5] M.DewarandD.Carlisle.Mathematicalsoftware: the
nextgeneration? In[1]
[6] M.Kohlhase.OMDoc: Towards aninternetstandardfor
theadministration,distributionandteachingof
mathematicalknowledge.InAISC'2000,2000.
[7] E.MelisandJ.H.Siekmann.Knowledge-basedproof
planning.ArticialIntelligence,115(1):65{105,1999.
[8] E.Melis, E.Andres,G.Goguadze,P.Libbrecht,M.
Pollet,andC.Ullrich.ActiveMath: System
description.InJohannaD.Moore,CarolRedeld,and
W.LewisJohnson,eds,ArticialIntelligencein
Education,pages580{582.IOSPress,2001.
[9] N.F.Noy,R.W.Fergerson,andM.A.Musen.The
knowledgemodelofProtege-2000: Combining
interoperabilityandexibility.In2thInternational
Conference onKnowledgeEngineeringandKnowledge
Management(EKAW'2000),2000.
[10] A.SorgatzandR.Hillebrand.MuPAD-Ein
ComputeralgebraSystemI.Linux Magazin,(12),1995.
[11] TheDublinCoreMetadataInitiative,1998,
http://purl.org/DC/.
13
TheAmericanMathematicalSociety'sE-Printservice,e.g.,
allowsthesearchthroughthereviewsonlywhicharetradi-
tionally encodedin the T
E
X language. Searching through
formulasinthis languageis oneof themost unpredictable