bootcat bootstrapping corpora and terms from the web. lrecnew

《bootcat bootstrapping corpora and terms from the web. lrecnew》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、BootCaT:BootstrappingCorporaandTermsfromtheWebMarcoBaroniandSilviaBernardiniSSLMIT,UniversityofBolognaCorsodellaRepubblica136,47100Forl`ı,Italyfbaroni,silviag@sslmit.unibo.itAbstractThispaperintroducestheBootCaTtoolkit,asuiteofperlprogramsimplementinganiterativeproceduretobootstrapspecializedcorp

2、oraandtermsfromtheweb.Theprocedurerequiresonlyasmallsetofseedtermsasinput.TheseedsareusedtobuildacorpusviaautomatedGooglequeries,andmoretermsareextractedfromthiscorpus.Inturn,thesenewtermsareusedasseedstobuildalargercorpusviaautomatedqueries,andsoforth.Thecorpusandtheunigramtermsarethenusedtoextr

3、actmulti-wordterms.WeconductedanevaluationofthetoolsbyapplyingthemtotheconstructionofEnglishandItaliancorporaandtermlistsfromthedomainofpsychiatry.Theresultsillustratethepotentialusefulnessofthetools.1.Introductiongrams,lookatintermediateoutputfiles,addnewtoolstothesuite,orchangeoneprogramwithouth

4、avingtoworryDespitecertainobviousdrawbacks(e.g.,lackofcon-abouttheothers.trol,sampling,documentationetc.),thereisnodoubtthattheWorldWideWebisamineoflanguagedataofunprece-Therestofthepaperisstructuredasfollows:In2.wedentedrichnessandeaseofaccess(KilgarriffandGrefen-shortlyreviewsomerelatedwork.In3

5、.wedescribethestette,2003).Itisalsotheonlyviablesourceof“dispos-algorithmimplementedintheBootCaTtools.In4.weable”corpora(Varantola2003)builtadhocforaspecificpresentanexperimentalevaluationofthetools.Wecon-purpose(e.g.,atranslationtask,thecompilationofater-cludein5.bysuggestingpossibledirectionsfor

6、furtherde-minologicaldatabase,domain-specificmachinelearning).velopmentandevaluationofthetoolkit.Thesecorporaareessentialresourcesforlanguagepro-fessionalswhoroutinelyworkwithspecializedlanguages,2.Relatedworkwherenewtermsareintroducedatafastpaceandstan-Theideaofbuildingacorpususingautomatedsearch

7、dardreferencecorporamustbecomplementedbyeasy-to-enginequeriesoriginatesfromGhanietal.(2001),whoconstruct,focused,up-to-datetextcollections.applyittothecreationofminoritylanguagecorpora.Whileitispossibletoconstructaweb-