What is a Knowledge Graph? | Ontotext Fundamentals

The heart of the knowledge graph is a knowledge model: a collection of interlinked descriptions of concepts, entities, relationships and events. Knowledge ... MakingSenseofTextandData Linkedin Twitter Youtube TextAnalysisforContentManagementConnectedInventory Interlinkyourorganization’sdataandcontentbyusingknowledgegraphpowerednaturallanguageprocessingwithourContentManagementsolutions. Features CaseStudies ShowcaseDemonstrators Learnmore… WhitePaperTextAnalysisforContentManagement ImplementaConnectedInventoryofenterprisedataassets,basedonaknowledgegraph,togetbusinessinsightsaboutthecurrentstatusandtrends,riskandopportunities,basedonaholisticinterrelatedviewofallenterpriseassets. TypicalDataAssets CaseStudies Learnmore… CaseStudyAKG-PoweredConnectedInventoryforaGlobalBank Sharethisarticle: Healthcare&LifeSciencesFinancialServicesIndustryMedia&PublishingPublicSector Quickandeasydiscoveryinclinicaltrials,medicalcodingofpatients’records,advanceddrugsafetyanalytics,knowledgegraphpowereddrugdiscovery,regulatoryintelligenceandmanymore BusinessApplications CaseStudies ShowcaseDemonstrators News,Events&BlogPosts Learnmore… TargetDiscoveryIdentifyNewDrugTargetsOrPromisingDrugRepurposingCandidatesQuicklyAndEasily Makebettersenseofenterprisedataandassetsforcompetitiveinvestmentmarketintelligence,efficientconnectedinventorymanagement,enhancedregulatorycomplianceandmore BusinessApplications CaseStudies ShowcaseDemonstrators News,Events&BlogPosts Learnmore… WhitepaperExploretheFinacialIndustryBusinessOntology(FIBO)withGraphDB Connectandmodelindustrysystemsandprocessesfordeeperdata-driveninsightsin: Manufacturing AutomotiveIndustry Energy BuildingAutomation Infrastructure Aerospace&Defense BusinessApplications CaseStudies ShowcaseDemonstrators News,Events&BlogPosts Learnmore… ShowcaseDemonstratorEnergyTransparencyKnowledgeGraph   Improveengagement,discoverabilityandpersonalizedrecommendationsforFinancialandBusinessMedia,MarketIntelligenceandInvestmentInformationAgencies,Science,TechnologyandMedicinePublishers,etc. BusinessApplications CaseStudies ShowcaseDemonstrators News,Events&BlogPosts Learnmore… WhitePaperSmarterContentwithaDynamicSemanticPublishingPlatform UnlockthepotentialfornewintelligentpublicservicesandapplicationsforGovernment,DefenceIntelligence,etc. BusinessApplications CaseStudies ClientsServices News,Events&BlogPosts Learnmore… CaseStudyUKParliament’sDataServiceArePoweredbyOntotext’sGraphDB Sharethisarticle: GraphDBOntotextPlatformMetadataStudio Linkdiversedata,indexitforsemanticsearchandenrichitviatextanalysistobuildbigknowledgegraphs. Learnmore… GetFreeEdition GetStandardEdition GetEnterpriseEdition ReleaseNotes QuickStartGuide Documentation Organizeyourinformationanddocumentsintoenterpriseknowledgegraphsandmakeyourdatamanagementandanalyticsworkinsynergy. Learnmore… RequestaLicense ReleaseNotes QuickStartGuide Documentation Integrateandevaluateanytextanalysisserviceonthemarketagainstyourowngroundtruthdatainauserfriendlyway. Learnmore… GetInTouch Documentation(upcoming) ReleaseNotes(upcoming) Sharethisarticle: KnowledgeHub Fundamentals WhatisaKnowledgeGraph? Theheartoftheknowledgegraphisaknowledgemodel:acollectionofinterlinkeddescriptionsofconcepts,entities,relationshipsandevents.Knowledgegraphsputdataincontextvialinkingandsemanticmetadataandthiswayprovideaframeworkfordataintegration,unification,analyticsandsharing. Theheartoftheknowledgegraphisaknowledgemodel–acollectionofinterlinkeddescriptionsofconcepts,entities,relationshipsandeventswhere: Descriptionshaveformalsemanticsthatallowbothpeopleandcomputerstoprocesstheminanefficientandunambiguousmanner; Descriptionscontributetooneanother,forminganetwork,whereeachentityrepresentspartofthedescriptionoftheentitiesrelatedtoit; Diversedataisconnectedanddescribedbysemanticmetadataaccordingtotheknowledgemodel. DoyouwanttolearnmoreaboutEnterpriseKnowledgeGraphs? KeyCharacteristics Knowledgegraphscombinecharacteristicsofseveraldatamanagementparadigms: Database,becausethedatacanbeexploredviastructuredqueries; Graph,becausetheycanbeanalyzedasanyothernetworkdatastructure; Knowledgebase,becausetheybearformalsemantics,whichcanbeusedtointerpretthedataandinfernewfacts. Knowledgegraphs,representedinRDF,providethebestframeworkfordataintegration,unification,linkingandreuse,becausetheycombine: Expressivity:ThestandardsintheSemanticWebstack–RDF(S)andOWL–allowforafluentrepresentationofvarioustypesofdataandcontent:dataschema,taxonomiesandvocabularies,allsortsofmetadata,referenceandmasterdata.TheRDF*extensionmakesiteasytomodelprovenanceandotherstructuredmetadata. Performance:Allthespecificationshavebeenthoughtout,andproveninpractice,toallowforefficientmanagementofgraphsof billionsoffactsandproperties. Interoperability:Thereisarangeofspecificationsfordataserialization,access(SPARQLProtocolforend-points),management(SPARQLGraphStore)andfederation.Theuseofgloballyuniqueidentifiersfacilitatesdataintegrationandpublishing. Standardization:AlltheaboveisstandardizedthroughtheW3Ccommunityprocess,tomakesurethattherequirementsofdifferentactorsaresatisfied–allthewayfromlogicianstoenterprisedatamanagementprofessionalsandsystemoperationsteams. Clickonimagetoenlarge OntologiesandFormalSemantics Ontologiesrepresentthebackboneoftheformalsemanticsofaknowledgegraph.Theycanbeseenasthedataschemaofthegraph.Theyserveasaformalcontractbetweenthedevelopersoftheknowledgegraphanditsusersregardingthemeaningofthedatainit.Ausercouldbeanotherhumanbeingorasoftwareapplicationthatwantstointerpretthedatainareliableandpreciseway.Ontologiesensureasharedunderstandingofthedataanditsmeanings. Whenformalsemanticsareusedtoexpressandinterpretthedataofaknowledgegraph,thereareanumberofrepresentationandmodelinginstruments: Classes.Mostoftenanentitydescriptioncontainsaclassificationoftheentitywithrespecttoaclasshierarchy.Forinstance,whendealingwithbusinessinformationtherecouldbeclassesPerson,OrganizationandLocation.PersonsandorganizationscanhaveacommonsuperclassAgent.Locationusuallyhasnumeroussub-classes,e.g.,Country,Populatedplace,City,etc.Thenotionofclassisborrowedbytheobject-orienteddesign,whereeachentityusuallybelongstoexactlyoneclass. Relationshiptypes.Therelationshipsbetweenentitiesareusuallytaggedwithtypes,whichprovideinformationaboutthenatureoftherelationship,e.g.,friend,relative,competitor,etc.Relationshiptypescanalsohaveformaldefinitions,e.g.,thatparent-ofisinverserelationofchild-of,theybotharespecialcasesofrelative-of,whichisasymmetricrelationship.Ordefiningthatsub-regionandsubsidiaryaretransitiverelationships. Categories.Anentitycanbeassociatedwithcategories,whichdescribesomeaspectofitssemantics,e.g.,“Bigfourconsultants”or“XIXcenturycomposers”.Abookcanbelongsimultaneouslytoallthesecategories:“BooksaboutAfrica”,“Bestseller”,“BooksbyItalianauthors”,“Booksforkids”,etc.Thecategoriesaredescribedandorderedintotaxonomy. Freetextdescriptions.Oftena‘human-friendlytext’descriptionisprovidedtofurtherclarifydesignintentionsfortheentityandimprovesearch. WhatisNOTaKnowledgeGraph? NoteveryRDFgraphisaknowledgegraph.Forinstance,asetofstatisticaldata,e.g.theGDPdataforcountries,representedinRDFisnotaKG.Agraphrepresentationofdataisoftenuseful,butitmightbeunnecessarytocapturethesemanticknowledgeofthedata.Itmightbesufficientforanapplicationtojusthaveastring‘Italy’associatedwiththestring‘GDP’andanumber‘1.95trillion’withoutneedingtodefinewhatcountriesareorwhatthe‘GrossDomesticProduct’ofacountryis.It’stheconnectionsandthegraphthatmaketheKG,notthelanguageusedtorepresentthedata. Noteveryknowledgebaseisaknowledgegraph.AkeyfeatureofaKGisthatentitydescriptionsshouldbeinterlinkedtooneanother.Thedefinitionofoneentityincludesanotherentity.Thislinkingishowthegraphforms.(e.g.AisB.BisC.ChasD.AhasD).Knowledgebaseswithoutformalstructureandsemantics,e.g.Q&A“knowledgebase”aboutasoftwareproduct,alsodonotrepresentaKG.Itispossibletohaveanexpertsystemthathasacollectionofdataorganizedinaformatthatisnotagraphbutusesautomateddeductiveprocessessuchasasetof‘if-then’rulestofacilitateanalysis. ExamplesofBigKnowledgeGraphs GoogleKnowledgeGraph.Googlemadethistermpopularwiththeannouncementofitsknowledgegraphin2012.However,thereareveryfewtechnicaldetailsaboutitsorganization,coverageandsize.TherearealsoverylimitedmeansforusingthisknowledgegraphoutsideGoogle’sownprojects. DBPedia.ThisprojectleveragesthestructureinherentintheinfoboxesofWikipediatocreateanenormousdatasetof4.58things(linkhttps://wiki.dbpedia.org/about)andanontologythathasencyclopediccoverageofentitiessuchaspeople,places,films,books,organizations,species,diseases,etc.ThisdatasetisattheheartoftheOpenLinkedDatamovement.Ithasbeeninvaluablefororganizationstobootstraptheirinternalknowledgegraphswithmillionsofcrowdsourcedentities. Geonames.Underacreativecommons,usersofGeonamesdatasethaveaccessto25milliongeographicalentitiesandfeatures. Wordnet.Oneofthemostwell-knownlexicaldatabasesfortheEnglishlanguage,providingdefinitionsandsynonyms.OftenusedtoenhancetheperformanceofNLPandsearchapplications. FactForge.Afteryearsofdevelopingexpertiseinthenewspublishingindustry,OntotextproducedtheirknowledgegraphofLinkedOpenDataandnewsarticlesaboutpeople,organizationsandlocations.ItincorporatesthedatafromtheKGsdescribedaboveaswellasspecializedontologiessuchastheFinancialIndustryBusinessOntology. KnowledgeGraphsandRDFDatabases Yearsago,wemovedawayfromthebuzzwordofBigDatatoSmartData.Havingunprecedentedamountsofdatapushedtheneedtohaveadatamodelthatmirroredourowncomplexunderstandingofinformation. Tomakedatasmart,themachinesneededtobenolongerboundbyinflexibledataschemasdefined‘apriori’.Weneededdatarepositoriesthatcouldrepresentthe‘realworld’andthetangledrelationshipsthatareentailed.Allthisneededtobedoneinamachine-readablewayandhaveaformalsemanticstoenableautomatedreasoningthatcomplementedandfacilitatedourown. RDFdatabases(alsocalledRDFtriplestores),suchasOntotext’sGraphDB,cansmoothlyintegrateheterogeneousdatafrommultiplesourcesandstorehundredsofbillionsoffactsaboutanyconceivableconcept.TheRDFgraphstructureisveryrobust(itcanhandlemassiveamountsofdataofallkindsandfromvarioussources)andflexible(itdoesnotneeditsschemare-definedeverytimeweaddnewdata). Aswehavealreadyseen,therearemanyfreelyavailableinterlinkedfactsfromsourcessuchasDBpedia,GeoNames,Wikidataandsoon,andtheirnumbercontinuestogroweveryday.However,therealpowerofknowledgegraphscomeswhenwetransformourowndataintoRDFtriplesandthenconnectourproprietaryknowledgetoopenglobalknowledge. AnotherimportantfeatureofRDFdatabasesistheirinferencecapabilitywherenewknowledgecanbecreatedfromalreadyexistingfacts.WhensuchnewfactsarematerializedandstoredinanRDFdatabase,oursearchresultsbecomemuchmorerelevant,openingnewavenuesforactionableinsights. Butifwewanttoaddevenmorepowertoourdata,wecanusetextminingtechniquestoextracttheimportantfactsfromfree-flowingtextsandthenaddthemtothefactsinourdatabase. HowCanKnowledgeGraphsHelpTextAnalysis Itisnosurprisethatmoderntextanalysistechnologymakesconsiderableuseofknowledgegraphs: Biggraphsprovidebackgroundknowledge,human-likeconceptandentityawareness,toenableamoreaccurateinterpretationofthetext; Theresultsoftheanalysisaresemantictags(annotations)thatlinkreferencesinthetexttospecificconceptsinthegraph.Thesetagsrepresentstructuredmetadatathatenablesbettersearchandfurtheranalytics; Factsextractedfromthetextcanbeaddedtoenrichtheknowledgegraph,whichmakesitismuchmorevaluableforanalysis,visualizationandreporting. OntotextPlatformimplementsallflavorsofthisinterplaylinkingtextandbigknowledgegraphstoenablesolutionsforcontenttagging,classificationandrecommendation.Itisaplatformfororganizingenterpriseknowledgeintoknowledgegraphs,whichconsistsofasetofdatabases,machinelearningalgorithms,APIsandtoolsforbuildingvarioussolutionsforspecificenterpriseneeds. OneinterestingexampleofsemantictaggingonnewsagainstabigknowledgegraphdevelopedaroundDBPediaisOntotext’sNOWpublicnewsservice. WhatAreKnowledgeGraphsUsedfor? Anumberofspecificusesandapplicationsrelyonknowledgegraphs.Examplesincludedataandinformation-heavyservicessuchasintelligentcontentandpackagereuse,responsiveandcontextuallyawarecontentrecommendation,knowledgegraphpowereddrugdiscovery,semanticsearch,investmentmarketintelligence,informationdiscoveryinregulatorydocuments,advanceddrugsafetyanalytics,etc.   Wanttolearnhowknowledgegraphshelpenterprisesimprovetheirknowledgemanagementandgetacompetitiveadvantage?   WhitePaper:KnowledgeGraphsintheEnterprise TheStoryBehindtheHype OntotextNewsletter