资源描述：
《An Open Distributed Architecture for Reuse and Integration of Heterogenous NLP Components》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。
1、AnOpenDistributedArchitectureforReuseandIntegrationofHeterogeneousNLPComponentsR4miZajac,MarkCasperandNigelSharpiesComputingResearchLaboratoryNew-MexicoStateUniversity{zajac,mcasper,nige1}©crl.nmsu.eduAbstractengineering:reuseandintegrationinthecontextofsoftwarearchitecture
2、sforNaturalLanguagePro-TheshiftfromComputationalLinguisticscessing.TheemphasisisputonreuseofNLPsoft-toLanguageEngineeringisindicativeofware,componentsandtheirintegrationinordertonewtrendsinNLP.Thispaperreviewsbuildlarge-scaleapplications.AlsorelevanttothistwoNLPengineeringp
3、roblems:reuseandpresentationaretopicssuchasintegrationofhetero-integration,whilerelatingtheseconcernsgeneouscomponentsforbuildinghybridsystemsortothelargercontextofappliedNLP.Itforintegratingspeechandother"higher-level"NLPpresentsasoftwarearchitecturewhichiscomponents(secti
4、on2).gearedtosupportthedevelopmentofava-Section3presentstheCorelliDocumentProcess-rietyoflarge-scaleNLPapplications:In-ingArchitecture,anewsoftwarearchitectureforformationRetrieval,CorpusProcessing,NLPwhichisdesignedtosupportthedevelopmentMultilingualMT,andintegrationofSpee
5、chofavarietyoflarge-scaleNLPapplications:Infor-Components.mationRetrieval,CorpusProcessing,MultilingualMT,andintegrationofSpeechwithotherNLPcom-ponents.1IntroductionTheshiftfromComputationalLinguisticstoLan-2ReuseinNLPguageEngineering1isindicativeofnewtrendsinThereisanincre
6、asingamountofsharedcorporaandNLP.Webelievethatitisnotsimplyanewfash-lexicalresourcesthatarebeingmadeavailableforionbutthatitisindicativeofthegrowingmatura-NLPresearchersthroughmanageddatarepositoriestionofthefield,asalsosuggestedbyanemphasisonsuchasLDC,CLR,ELRA,etc.(seee.g.
7、,(Wilksbuildinglarge-scalesystems,awayfromtoyresearchetal.92)foranoverviewoftheserepositories).systems.ThereisalsoanincreasingawarenessthatTheseresourcesconstitutethebasicrawmaterialsreal-sizesystemsarenotmerescaled-uptoysystemsforbuildingNLPsoftwarebutnotallofthesere-butth
8、attheypresentanaltogetherqualitativelydif-sourcescanbereadilyused:theymightbeavail
