Southampton and The Open University. Preface(15)

Advanced Knowledge Technologies is recognised as a leading research programme conducted at some of the foremost informatics departments in Britain. It is also a training ground for a new generation of researchers. To highlight the work of these students, a

di eringreliability;informationfoundcanevenbespam,typos,deliberatemis-

informationorsimplyerroneous.Wheninformationiscontainedintextualdoc-

uments,extractingitrequiresmoresophisticatedIEmethodologiesbasedon

linguisticanalysisandmethodstoensureadegreeofreliabilityofextracted

information.

Informationfromwhateversourceisoftenredundant,i.e.thatitcanbe

foundindi erentcontextsandindi erentsuper cialformats-theredundancy

ofinformationcaninitselfbeaweakproofofitsvalidity[5].However,wesuggest,

thatthisaloneisnotenough,additionalinformationeitherwithinorsurrounding

extractedentitiescanbeusedtodevelopadditionalevidence.Armadillonowuses

anevidencebuildingapproachofnumerousrudimentarytechniques,(basedon

SimMetrics,Section3).

4.2HowArmadilloworks?

Armadillolearnshowtobestextractinformationinthefollowingway:

1.itminesacoherentportionoftherepository(e.g.awebsiteoraclassof

sites);

2.itintegratesinformationandassignsreliabilitiesofdi erentsources(e.g.

digitallibraries,services,webpages).Theseratingsareusedtodirectthe

learningfromtherepository;

3.itdiscoversnewinformationintherepositorythatinturnisratedandused

tobootstrapnewlearninguntilastableinformationbaseisreached;

4.itstorestheharvestedinformationintoaRDFKnowledgebase.Thedatabase

canthenbeusedtoaccesstheextractedinformation(asdetailedlater)or

toproduceindicesfordocumentretrievalorannotation.

Armadilloisadatadrivensystemtypicallybeginingfromrigidlystructured

reliablesourcesusingexamplesprovidedbyeitherawrapper,theuserorprevious

data.Armadillousespreviouslyobtainedseedstolearnonmorecomplexsources

(e.g.freetexts)usingthepreviouslyacquiredinformation.

InordertoexplaininmoredepththeworkingofArmadilloanexampleis

nowdetailed.

4.3TheComputerScienceDepartmentApplication

ConsiderthefollowingexampletaskofminingwebsitesofComputerScience

Departmentsto ndacademics(name,position,homepage,emailaddress,tele-

phonenumberandalistofpublicationsmorecompletethantheoneprovided

byrepositoriessuchasCiteseer).

Simplydiscoveringwhoworksforadepartmentismorecomplexthangeneric

NamedEntityRecognition(NER)asmanyirrelevantpeople’snamesaremen-

tionedinasite,sofundergraduatestudents,secretaries,aswellas

namesofresearchersfromexternalsitesandhenceirrelevantforthistask.

Armadillousesastatisticalevidencebasedloopingapproachtoaidtheval-

idationtask.Initiallyaquicklistofpotentialnamesofpeopleworkinginthe

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(15)全文阅读和word下载服务。