Southampton and The Open University. Preface(17)

morepotentialknowledgewhichmustthenbevalidated.

EvidenceBuildingandValidation-comesfromtheseriesofweakap-

proacheseachofwhicharecombinedtogivearatingwhichcanbeusedto

validateknowledge.Firstlyadditionalredundantsourcesofthesamedataare

identi edviaasimplesearch,e.g.google ndingrelevantURLs.Thisredun-

dancybyitselfisnolongerconsideredvalidevidenceasthereliabilityofsources

themselvesmustalsobetakenintoaccount,e.g.asourcewithlotsofpoten-

tialacademicsisconsideredabettersourcethanasourcedetailingjustasingle

academic.

Theratingofsourcesin uencesArmadillo’sperceivedvalidityofthepoten-

ingthepreviouslymentionedSimMetriclibrary,section

3,theextractedentitiesarecrossexaminedforsimplesimilarities:ifsomething

issuitablydissimilarfromtherest,forexampleafailedcapture40timeslonger

thannormal,itisconsideredmorepossiblyanerroranditsvalidityratingis

decreasedaswellastheratingofthesourcefromwhichitwasfound.Atpresent

thecrosssimilaritytestsareacombinationofSimMetric’ssimilaritymetrics,

althoughmorecomplexapproachescouldbeemployed.Thecontextarounda

capturecanalsodetaillikelihood,e.g.thesimilarityofPartofSpeechtagssur-

roundinginstances,variousothersimilaritytechniquestoprovidefurtherweak

evidencecouldbeused,forexampleavectorspacemodelofthesource(e.g.web-

page)similarity,acomparisonofsimilaritiesinalinkanalysisofdatasourcesor

ananalysisofthesourcedocumentsDOMstructureanditssimilaritytoother

sources.

Thiscombinationofmultipleweaktechniquescanprovideimprovedcon-

denceintheextractedknowledgeArmadillo nds,(againalthoughnotyet

implementeditisenvisionedtoprioritisetechniquesthroughInformationGain

andmemoryortimecosts).

Extractionofpotentialknowledge-armadilloasinpreviouspapers[5]inte-

gratestheAmilcare[2](LP)2algorithmtoextractpotentialnewknowledge,but

canbeextendedtoencompassdi erentMLalgorithms,forexampleT-Rex[7].

Usingratingsfromtheevidencebuildingapproach(Section4.3)theInformation

Extractionlearnsitscontextualrulesonthemosthighlyratedsourcesusingthe

mostlikelyinstancesasseeddata.

搜索“diyifanwen.net”或“第一范文网”即可找到本站免费阅读全部范文。收藏本站方便下次阅读，第一范文网，提供最新人文社科Southampton and The Open University. Preface(17)全文阅读和word下载服务。