挖掘技术教程_第1页
挖掘技术教程_第2页
挖掘技术教程_第3页
挖掘技术教程_第4页
挖掘技术教程_第5页
已阅读5页,还剩52页未读 继续免费阅读

下载本文档

版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领

文档简介

第十三章半第十三章半结构化文本挖杨建北京大学计算机科学技术研究1Text-centricXMLText-centricXMLDocumentsmarkedupasE.g.,assemblymanuals,journalQueriesareuserinformationE.g.,givemetheSection(element)ofdocumentthattellsmehowtochangeabrakelight2VectorspacesandVectorspacesandVectorspaces–tried+testedframeworkforkeywordretrievalOther“bagofwords”applicationsinclassification,clustering…Fortext-centricXMLretrieval,canwemakeuseofvectorspaceideas?Challenge:capturethestructureofanXMLdocumentinthevectorspace.3VectorspacesandForinstance,distinguishbetweenVectorspacesandForinstance,distinguishbetweenfollowingtwoBillBill4Content-richXML:Content-richXML:Lexicon5EncodingtheGatesEncodingtheGatesWhataretheaxesofthevectorIntextretrieval,therewouldbeasingleaxisforGatesHerewemustseparateoutthetwooccurrences,underAuthorandTitleThus,axesmustrepresentnotonlyterms,butsomethingabouttheirpositioninanXMLtree6Beforeaddressingthis,letustheBeforeaddressingthis,letusthekindsofquerieswewantto7SubtreesandConsiderallsubtreesofthethatincludeatleastonelexiconSubtreesandConsiderallsubtreesofthethatincludeatleastonelexicon…8Structuralterms:Calleachoftheresulting(8+,inthepreviousslide)subtreesastructuralStructuralterms:Calleachoftheresulting(8+,inthepreviousslide)subtreesastructuraltermCreateoneaxisinthevectorspaceforeachdistinctstructuraltermEachdocumentbecomesavectorinthespaceofstructuraltermsAquerytreecanlikewisebefactoredintostructuraltermsAndrepresentedasaAllowsweightingportionsofthe9StructuraltermsStructuraltermsWeightsbasedonfrequenciesfornumberofoccurrences(justaswehadtf)Alltheusualissueswithterms(stemming?Casefolding?)remainExampleoftfTobeExampleoftfTobeornottoExercise:HowmanyaxesarethereinthisHerethestructuraltermscontainingtoorbewouldhavemoreweightthanthosethatForthedocontheleft:inastructuraltermrootedatthenodePlay,shouldn’tForthedocontheleft:inastructuraltermrootedatthenodePlay,shouldn’tHamlethaveahighertfweightthanIdea:multiplytfcontributionofatermtoanodeklevelsupbyk,forsomeg<Alaspoor=0.8Forthedoconthepreviousslide,theHamletismultipliedbyYorickismultipliedbyinanystructuraltermrootedatThenumberofThenumberofstructuralCanbeAlright,howhuge,ImpracticaltobuildavectorindexwithsomanyWillexaminepragmaticsolutionstothisshortly;fornow,continuetobelieve…RestrictstructuralRestrictstructuralDependingontheapplication,wemayrestrictthestructuraltermsE.g.,mayneverwanttoreturnaTitlenode,onlyBookorPlaynodesSodon’tenumerate/index/retrieve/scorestructuraltermsrootedatsomenodesTwoQuery-timematerializationofRestrictthekindsofsubtreestoamanageableQuery-timeInsteadofenumeratingallstructuraltermsofalldocsQuery-timeInsteadofenumeratingallstructuraltermsofalldocs(andthequery),enumerateonlyforthequeryThelatterishopefullyasmallNow,we’rereducedtocheckingwhichstructuralterm(s)fromthequerymatchasubtreeofanyThisistreepatternmatching:givenatexttreeandapatterntree,findmatchesExceptwehavemanytextOurtreesarelabeledandTextHereweseekadocwithHamletintheTextHereweseekadocwithHamletintheOnfindingthematchwecomputethecosinesimilarityscoreAfterallmatchesarefound,rankbysortingAlaspoorQuery(StillAdoc(StillAdocwithYoricksomewhereinQueryWillgettoitRestrictingtheRestrictingtheEnumeratingallstructuralterms(subtrees)isprohibitive,forindexingMostsubtreesmayneverbeusedinprocessinganyqueryCanwegetawaywithindexingarestrictedclassofsubtreesIdeally–focusonsubtreeslikelytoariseinJuruXML(IBMOnlypathsincludingalexicontermInJuruXML(IBMOnlypathsincludingalexicontermInthisexamplethereareonly14(why?)suchpathsThuswehave14structuraltermsintheTobeornottoWhyisthisfarmoreHowbigcantheindexbeasafunctionoftheExampleofaretrievalST=StructuralExampleofaretrievalST=StructuralNowranktheDoc’sbycosinesimilarity;e.g.,Doc9scores0.578.SQLforSQLforUsageHuman-readableData-orientedMixeddocuments(e.g.,patientReliesXMLSchemaTheprincipalformsTheprincipalformsofXQueryexpressionsare:pathelementFLWR("flower")listconditionalquantifieddatatypeEvaluatedwithrespecttoaFOR$pINdocument("bib.xml")//publisherLETFOR$pINdocument("bib.xml")//publisherLET$b:=document("bib.xml”)//book[publisher=$p]WHEREcount($b)>100RETURN$pFORgeneratesanorderedlistofbindingsofpublishernamesto$pLETassociatestoeachbindingafurtherbindingofthelistofbookelementswiththatpublisherto$batthisstage,wehaveanorderedlistoftuplesofbindings:WHEREfiltersthatlisttoretainonlythedesiredRETURNconstructsforeachtuplearesultingQueriesSupportedbyQueriesSupportedbyLocation/position(“chapterSimple/play/titlecontainsPathtitlecontains/play//titlecontainsComplexEmployeeswithtwoSubsumes:WhataboutrelevanceXQuery:OrderXQuery:OrderByOrderbyclauseonlyallowsorderingbySaybyanattributeRelevanceIsoftenCan’tbeexpressedeasilyasfunctionofsettobeIsbetterabstractedoutofqueryformulation(cf.XQuery:OrderByXQuery:OrderByfor$dinlet$e:=document("emps.xml")//emp[deptno=$d]wherecount($e)>=10orderbyavg($e/salary)descendingreturn<big-dept>{$d,XMLXMLNativeXMLNativeXMLUsesXMLdocumentaslogicalShouldPCDATA(parsedcharacterDocumentContrastDBmodifiedforGenericIRsystemmodifiedforXMLIndexingandMostnativeXMLIndexingandMostnativeXMLdatabasestakenaDBNoIRtyperelevanceOnlyafewthatfocusonrelevanceDatavs.Text-centricDatavs.Text-centricmainlyarecastingofrelationalusedformessagingbetweenenterpriseusedforannotatingRichinDemandsgoodintegrationoftextretrievalE.g.,findmetheISBN#sofBookswithatleastthreeChaptersdiscussingcocoaproduction,rankedbyPriceDatastructuresDatastructuresforXMLDatastructuresforDatastructuresforXMLWhataretheprimitivesweInvertedindex:givemeallelementsmatchingtextqueryQWeknowhowtodothis–treateachelementasadocumentGivemeallelements(immediately)belowanyinstanceoftheBookelementCombinationoftheParent/childNumbereachParent/childNumbereachMaintainalistofparent-childE.g.,Chapter:21EnablesimmediateButwhatabout“thewordHamletunderSceneelementunderaPlayGeneralpositionalViewtheXMLdocumentasGeneralpositionalViewtheXMLdocumentasatextBuildapositionalindexforeachMarkthebeginningandendforeachelement,PositionaldroppethunderVersePositionaldroppethunderVerseunderPl3y.SummaryofdataSummaryofdataPathcontainmentetc.canessentiallybesolvedbypositionalinvertedindexesRetrievalconsistsof“merging”Allthecompressiontricksetc.from276AarestillComplicationsarisefrominsertion/deletionofelements,textwithinelementsBeyondthescopeofthisINEX:aINEX:abenchmarkfortext-XMLINitiativefortheINitiativefortheEvaluationofXMLrmatik.uni-SponsoredDELOSNetworkofExcellenceforDigitalIEEEComputerBenchmarkfortheevaluationofXMLConsistsSetofXMLCollectionofretrievalDocumentcollectioninterlinkingtheWiki)INEX--adhocINEX--adhocEachengineindexesEngineteamconvertsretrievaltasksintoInXMLquerylanguageunderstoodbyInresponse,theengineretrievesnotdocs,butelementswithindocsEngineranksretrievedINEXINEX2006INEXINEX2006(IEEEComputerSociety12,107494Averagearticle:1,532XMLAveragenodedepth=INEX2007(WikipediatheXMLfull-textsof659,388morethan60GB(4.6GBwithout30millionarticle:161.35depth:INEXEachINEXEachtopicisaninformationneed,oneoftwokinds:ContentOnly(CO)–freetextContentandStructure(CAS)–explicitstructuralconstraints,e.g.,containmentSampleINEXCOSampleINEXCO<Title>computationalbiology<Keywords>computationalbiology,bioinformatics,genome,genomics,proteomics,sequencing,proteinfolding<Description>Challengesthatarise,andapproachesbeingexplored,intheinterdisciplinaryfieldofcomputational<Narrative>Toberelevant,adocument/componentmusteithertalkingeneraltermsabouttheopportunitiesattheintersectionofcomputerscienceandbiology,ordescribeaparticularproblemandthewaysitisbeingattacked.</Narrative>INEXEachINEXEachengineformulatesthetopicasaE.g.,usethekeywordslistedintheEngineretrievesoneormoreelementsandranksthem.Humanevaluatorsassigntoeachretrievedelementrelevanceandcoveragescores.INEXhowINEXhowrelevantistheretrievedistheretrievedelementtoospecific,toogeneral,orjustrightE.g.,ifthequeryseeksadefinitionoftheFastFourierTransform,doIgettheequation(toospecific),thec

温馨提示

  • 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
  • 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
  • 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
  • 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
  • 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
  • 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
  • 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。

评论

0/150

提交评论