Abstract Corpus-based Sublanguage Modelling for NLP Applications a Tentative Methodology

《Abstract Corpus-based Sublanguage Modelling for NLP Applications a Tentative Methodology》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、Corpus-basedSublanguageModellingforNLPApplications:aTentativeMethodologyGuyDevilleFacultésUniversitaires,61ruedeBruxelles–B-5000Namur–BelgiqueAbstractThepresentpaperproposesamethodologyfornaturallanguagemodellingintheprospectofautomation.Theauthoradvocatesthat

2、anyenterpriseaimingattheformalisationofnaturallanguage(atanylevelofdescription)requiresathreefoldvalidation:(i)thedevelopedlanguagemodels(orlingwares)accountforlanguageproductionsrealizedinareal-scalecontext(empiricalvalidation).Atthisstagethecollationandexplo

3、itationoflanguagecorporaisakeyfactor;(ii)themodelsaimatmaximaldescriptiveadequacy:theyaremoremotivatedbylinguisticcriteriathanbytheirinternalcoherence(theoreticalvalidation);(iii)themodelsaretobeimplementedwithinthecurrentlimitationsofinformationtechnology(ope

4、rationalvalidation).Itappearshoweverthatmodelsthatfullysatisfyempirical,theoreticalandoperationalvalidityareimpossible,assuchmodelsreducethelanguagetoitstractabledimensions,andastheintractabilityofnaturallanguageingeneralhasbeenproved.Asaconsequence,computersc

5、anprocessspecificportionsofnaturallanguageonlyadequatelyinlimitedandwell-definedconceptualworlds.Whenreferringtothisspecializedknowledge,wewillshowthattheabovevaliditycriteriacanbesuccessfullyrevisitedinthelightofthenotionofsublanguage(withprototypicaldimensio

6、nsatlexical,syntactic,semanticandpragmaticlevels).Theapproachwillbeillustratedwithapplicationsinthefieldofmedicallanguage.1.IntroductionComputationallinguistics,alsolabelledlanguageengineering,humanlanguagetechnologyornaturallanguageprocessing(henceforthNLP)is

7、adisciplinethatissituatedatthecrossroadsoflanguagesciences,computerscience(includingA.I.)andcognitivesciences.Takeninitsbroadmeaning,languageengineeringconcernscomputerapplicationswithacomponentthatprocessnaturallanguage.Thisdefinitionencompassesalargevarietyo

8、factivitiesofheterogeneousnature,fromconcordancesoftwarepackagestoadvancedvoice-drivenhuman-machineinterfaces,throughfull-fledgedcomputationaltheoriesoflanguage.Soweneedtodefinemor