3G搜索 自然语义语言 整合统一处理 NLP 项目级实施.docx
- 文档编号:15371174
- 上传时间:2023-07-03
- 格式:DOCX
- 页数:13
- 大小:24.73KB
3G搜索 自然语义语言 整合统一处理 NLP 项目级实施.docx
《3G搜索 自然语义语言 整合统一处理 NLP 项目级实施.docx》由会员分享,可在线阅读,更多相关《3G搜索 自然语义语言 整合统一处理 NLP 项目级实施.docx(13页珍藏版)》请在冰点文库上搜索。
3G搜索自然语义语言整合统一处理NLP项目级实施
StructuralandsemanticaspectsofsimilarityofDocumentTypeDefinitionsandXMLschemas
ThenaturaloptimizationstrategyforXML-to-relationalmappingmethodsisexploitationofsimilarityofXMLdata.However,noneofthecurrentsimilarityevaluationapproachesissuitableforthispurpose.WhilethekeyemphasisiscurrentlyputonsemanticsimilarityofXMLdata,themainaspectofXML-to-relationalmappingmethodsisanalysisoftheirstructure.
Inthispaperweproposeanapproachthatutilizesaverifiedstrategyforstructuralsimilarityevaluation–treeeditdistance–toDTDconstructs.ThisapproachisabletocopewiththefactthatDTDsinvolveseveraltypesofnodesandcanformgeneralgraphs.Inaddition,itisoptimizedforthespecificfeaturesofXMLdataand,ifrequired,itenablesonetoexploitthesemanticsofelement/attributenames.Usingasetofexperimentsweshowtheimpactoftheseextensionsonsimilarityevaluation.And,finally,wediscusshowthisapproachcanbeextendedforXSDs,whichinvolveplentyof“syntacticsugar”,i.e.constructsthatarestructurallyorsemanticallyequivalent.
Aggregatedsearchofdataandservices
InformationSystems
Researchhighlights
►Fromauserperspective,dataandservicesarecomplementaryviewsofaninformationsource;►Aframeworktoeffectivelyaccessdataandservicesinaunifiedmannerisproposedandevaluated;►Theframeworkreliesonontologiesfordescribingbothdatasourcesandwebservices;►MappingsandIRtechniqueslocaterelevantservicesrelatedtoaqueryoverthedata.
Fromauserperspective,dataandservicesprovideacomplementaryviewofaninformationsource:
dataprovidedetailedinformationaboutspecificneeds,whileservicesexecuteprocessesinvolvingdataandreturninganinformativeresultaswell.Forthisreason,usersneedtoperformaggregatedsearchestoidentifynotonlyrelevantdata,butalsoservicesabletooperateonthem.Atthecurrentstateoftheartsuchaggregatedsearchcanbeonlymanuallyperformedbyexpertusers,whofirstidentifyrelevantdata,andthenidentifyexistingrelevantservices.
Inthispaperweproposeasemanticapproachtoperformaggregatedsearchofdataandservices.Inparticular,wedefineatechniquethat,onthebasisofanontologicalrepresentationofbothdataandservicesrelatedtoadomain,supportsthetranslationofadataqueryintoaservicediscoveryprocess.
Inordertoevaluateourapproach,wedevelopedaprototypethatcombinesadataintegrationsystemwithanovelinformationretrieval-basedWebServicediscoveryengine(XIRE).TheresultsproducedbyawidesetofexperimentsshowtheeffectivenessofourapproachwithrespecttoIRapproaches,especiallywhenWebServicedescriptionsareexpressedbymeansofaheterogeneousterminology.
Naturallanguagequeryingforvideodatabases
InformationSciences
Thevideodatabaseshavebecomepopularinvariousareasduetotherecentadvancesintechnology.Videoarchivesystemsneeduser-friendlyinterfacestoretrievevideoframes.Inthispaper,auserinterfacebasedonnaturallanguageprocessing(NLP)toavideodatabasesystemisdescribed.Thevideodatabaseisbasedonacontent-basedspatio-temporalvideodatamodel.Thedatamodelisfocusedonthesemanticcontentwhichincludesobjects,activities,andspatialpropertiesofobjects.Spatio-temporalrelationshipsbetweenvideoobjectsandalsotrajectoriesofmovingobjectscanbequeriedwiththisdatamodel.Inthisvideodatabasesystem,anaturallanguageinterfaceenablesflexiblequerying.Thequeries,whicharegivenasEnglishsentences,areparsedusinglinkparser.Thesemanticrepresentationsofthequeriesareextractedfromtheirsyntacticstructuresusinginformationextractiontechniques.Theextractedsemanticrepresentationsareusedtocalltherelatedpartsoftheunderlyingvideodatabasesystemtoreturntheresultsofthequeries.Notonlyexactmatchesbutsimilarobjectsandactivitiesarealsoreturnedfromthedatabasewiththehelpoftheconceptualontologymodule.Thismoduleisimplementedusingadistance-basedmethodofsemanticsimilaritysearchonthesemanticdomain-independentontology,WordNet.
Introduction
2.Relatedworkonnaturallanguagequeryprocessing
2.1.Naturallanguagequeryingoverdatabases
2.2.Naturallanguagetechniquesovervideodatabases
3.Videodatamodel
4.Queryprocessing
4.1.Semanticrepresentationsofqueries
4.2.Parsingqueries
4.3.Informationextractionmodule
5.Ontology-basedquerying
5.1.WordNetontology
5.2.Measuringsemanticsimilaritybetweenwords
5.3.Expandingsemanticrepresentationswithontology
6.Evaluation
7.Conclusion
Acknowledgements
Aschemaandontology-aidedintelligentinformationintegration
Theresearchissuesofintelligentinformationintegrationhavebecomeubiquitousandcriticallyimportantine-business(EB)withtheincreasingdependenceonInternet/Intranetandinformationtechnology(IT).Accessingtheintelligentinformationsourcesseparatelywithoutintegrationmayleadtothechaosofinformationrequested.Itisalsonotcost-effectiveinEBsettings.AcommongeneralwaytodealwithheterogeneityproblemsintraditionalIIIistocreateacommondatamodel.TheeXtensibleMarkupLanguage(XML)hasbeenthestandarddatadocumentformatforexchanginginformationontheWeb.XMLonlydealswiththestructuralheterogeneity;itcanbarelyhandlethesemanticheterogeneity.Ontologiesareregardedasanimportantandnaturalmeanstorepresenttheimplicitsemanticsandrelationshipsintherealworld.AndtheyareusedtoassisttoreachsemanticinteroperabilityinIIIinthisresearch.Inthispaper,weprovideagenericconstructorientationnoadhocmethodtogeneratetheglobalschematoenabletheweb-basedalternativetotraditionalIII.Weprovideawiserquerymethodovermultipleintelligentinformationsourcesbyapplyingglobal-as-view(GAV)andlocal-as-view(LAV)approachwiththeuseofontologytoenhancebothstructuralandsemanticinteroperabilityoftheunderlyingintelligentinformationsources.Weconstructaprototypeimplementingthemethodtoprovideaproofonthevalidityandfeasibility.
Alabeled-treeapproachtosemanticandstructuraldatainteroperabilityappliedinhydrologydomain
InformationSciences
Theissuesofdataintegrationandinteroperabilityposesignificantchallengesinscientifichydrologicalandenvironmentalstudies,duelargelytotheinherentsemanticandstructuralheterogeneitiesofmassivedatasetsandnon-uniformautonomousdatasources.Toaddressthesedataintegrationchallenges,weproposeaunifieddataintegrationframework,calledHydrologicalIntegratedDataEnvironment(HIDE).HIDEisbasedonalabeled-treedataintegrationmodelreferredtoasDataNodetree.Usingthisframework,characteristicsofdatasetsgatheredfromdiversedatasources–withdifferentlogicalandaccessorganizations–canbeextractedandclassifiedasTime–Space–Attribute(TSA)labelsandaresubsequentlyarrangedinaDataNodetree.Theuniquenessofourapproachisthatiteffectivelycombinesthesemanticaspectsofthescientificdomainwithdiversedatasetshavingdifferentlogicalorganizationstoformaunifiedview.Further,wealsoadoptametadata-basedapproachforspecifyingtheTSA-DataNodetreeinordertoachieveflexibilityandextensibility.ThesearchengineofourHIDEprototypesystemevaluatesasimpleuserquerysystematicallyontheTSA-DataNodetree,presentingintegratedresultsinastandardizedformatthatfacilitatesbotheffectiveandefficientdataintegration.
IntegrationofMultipleClassifiersforChineseSemanticDependencyAnalysis
ElectronicNotesinTheoreticalComputerScience
SemanticDependencyAnalysis(SDA)hasextensiveapplicationsinNaturalLanguageProcessing(NLP).Inthispaper,anintegrationofmultipleclassifiersispresentedforSDAofChinese.ANaiveBayesianClassifier,aDecisionTreeandaMaximumEntropyclassifierareusedinamajoritywinsvotingscheme.AportionofthePennChineseTreebankwasmanuallyannotatedwithsemanticdependencystructure.Theneachofthethreeclassifierswastrainedonthesametrainingdata.Allthreeoftheclassifierswereusedtoproducecandidaterelationsfortestdataandthecandidaterelationthathadthemajorityvotewaschosen.Theproposedapproachachievedanaccuracyof86%inexperimentation,whichshowsthattheproposedapproachisapromisingoneforsemanticdependencyanalysisofChinese.
Semanticinformationintegrationandquestionansweringbasedonpervasiveagentontology
ExpertSystemswithApplications
Thetraditionalsearchenginesreturnalargenumberofrelativewebpagesratherthanaccurateanswers.However,inaquestionansweringsystem,userscouldusesentencesindailylifetoraisequestions.Thequestionansweringsystemwillanalyzeandcomprehendthesequestionsandreturnanswerstousersdirectly.Aimingattheproblemsincurrentnetworkenvironment,suchaslowprecisionofquestionanswering,imperfectexpressionofdomainknowledge,lowreuserateandlackofreasonabletheoryreferencemodels,weputforwardtheinformationintegrationmethodofsemanticwebbasedonpervasiveagentontology(SWPAO)method,whichwillintegrate,analyzeandprocessenormouswebinformationandextractanswersonthebasisofsemantics.WithSWPAOmethodastheclue,wemainlystudythemethodofconceptextractionbasedonuniformsemantictermmining,pervasiveagentontologyconstructionmethodonaccountofmulti-pointsandtheanswerextractioninviewofsemanticinference.Meanwhile,wepresentthestructuralmodelofthequestionansweringsystemapp
- 配套讲稿:
如PPT文件的首页显示word图标,表示该PPT已包含配套word讲稿。双击word图标可打开word文档。
- 特殊限制:
部分文档作品中含有的国旗、国徽等图片,仅作为作品整体效果示例展示,禁止商用。设计者仅对作品中独创性部分享有著作权。
- 关 键 词:
- 3G搜索 自然语义语言 整合统一处理 NLP 项目级实施 3G 搜索 自然 语义 语言 整合 统一 处理 项目 实施