资源描述:
《NLP-Based Requirements Modeling Experiments on the Quality of the models》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、UNIVERSITYOFTRENTODEPARTMENTOFINFORMATIONANDCOMMUNICATIONTECHNOLOGY38050Povo–Trento(Italy),ViaSommarive14http://www.dit.unitn.itNLP-BASEDREQUIREMENTSMODELING:EXPERIMENTSONTHEQUALITYOFTHEMODELSNadzeyaKiyavitskaya,NicolaZeni,LuisaMichandJohnMylopoulosJanuary2004TechnicalReport#DIT-04-00
2、5.NLP-BasedRequirementsModeling:*ExperimentsontheQualityofthemodels1112NadzeyaKiyavitskaya,NicolaZeni,LuisaMich,JohnMylopoulos1DepartmentofInformationandTelecommunicationTechnologies,UniversityofTrento{nadzeya.kiyavitskaya,nicola.zeni,luisa.mich}@unitn.it2DepartmentofComputerScience,U
3、niversityofTorontojm@cs.toronto.eduAbstractConceptualmodelsareusedinavarietyofareaswithinComputerScience,includingSoftwareEngineering,DatabasesandAI.Amajorbottleneckinbroadeningtheirapplicabilityisthetimeittakestobuildaconceptualmodelforanewapplication.Notsurprisingly,avarietyoftoolsa
4、ndtechniqueshavebeenproposedforreusingconceptualmodels,e.g.ontologies,orforbuildingthemsemi-automaticallyfromnaturallanguage(NL)descriptions.Whathasbeenleftlargelyunexploredistheimpactofsuchtoolsonthequalityofthemodelsthatarebeingcreated.Thispaperpresentstheresultsofthreeexperimentsde
5、signedtoassesstheextenttowhichaNatural-LanguageProcessing(NLP)toolimprovesthequalityofconceptualmodels,specificallyobject-orientedones.OurmainexperimentalhypothesisisthatthequalityofadomainclassmodelishigherifitsdevelopmentissupportedbyaNLPsystem.Thetoolusedfortheexperiment–namedNL-OO
6、PS–extractsclassesandassociationsfromaknowledgebaserealizedbyadeepsemanticanalysisofasampletext.Specifically,NL-OOPSproducesclassmodelsatdifferentlevelsofdetailbyexploitingclasshierarchiesintheknowledgebaseofaNLPsystemandmarksambiguitiesinthetext.Inourexperiments,wehadgroupsworkingwit
7、handwithoutthetool,andthencomparedandevaluatedthefinalclassmodelstheyproduced.Theresultsoftheexperiments–thefirstonthistopic–giveinsightsonthestateoftheartoflinguistics-basedComputerAidedSoftwareEngineering(CASE)toolsandallowidentifyingimportantguidelinestoimprovetheirperformance.Inpa
8、rticu