1 Peering down the memory hole: history, censorship and the digital turn Glenn Tiffert 1 [email protected] Economic, technological and political forces are fundamentally reshaping how knowledge is produced and consumed. Much as clay tablets and bone gave way to papyrus and parchment, paper is now yielding to bitstreams. In turn, libraries are rebalancing their collections away from physical volumes towards online resources, where a small number of remote providers dominate storage and delivery, and effectively govern the terms of access for all. For many, printed books and periodicals are fast becoming reminders of things past. 2 This digital turn is seductive; it promises to deliver abundant knowledge with unprecedented convenience and immediacy. Yet, commentators have noted that it also concentrates power, exacerbates inequalities, and raises thorny theoretical, ethical, and methodological concerns. 3 For instance, a handful of media conglomerates now control more than half of all academic publishing, which makes it highly vulnerable to rent-seeking, poor stewardship, or deliberate manipulation. The oligopolistic pricing power publishers enjoy, protected by copyright, delivers enviable profit margins while squeezing even the richest institutions. 4 Though the questions are many, censorship has scarcely registered among them. Arguably, this reflects the privileged position of those who dominate the conversation: scholars who study and work in comparatively liberal societies, where one can reasonably assume that a publication accessible today will remain so tomorrow, allowing for normal conservation. This study dispels that innocence. Using a case from the People’s Republic of China (PRC), it explores some of the most disturbing ramifications of our deepening digital dependence. Simply put, the Chinese government is leveraging technology to quietly export its domestic 1 Glenn Tiffert is the Lieberthal-Rogel Center for Chinese Studies Distinguished Postdoctoral Fellow in Residence at the University of Michigan, Ann Arbor. The Lieberthal-Rogel Center for Chinese Studies generously funded this study. At the University of Michigan, Professors Mary Gallagher, Steven Abney, and Kerby Shedden offered valuable feedback. Fu Liangyu, the university’s Chinese Studies librarian, helped to locate and acquire materials. Luo Fusheng, Arden Shapiro, Yan Wei, and Yin Shiyuan provided essential research assistance. Margaret Orton coded the naïve Bayes classifier. 2 Watanabe, Teresa. “Universities Redesign Libraries for the 21st Century: Fewer Books, More Space.” Los Angeles Times (2017): Accessed July 26, 2017. http://www.latimes.com/local/lanow/la-me-college-libraries-20170419- story.html. 3 Roy Rosenzweig, “Scarcity or Abundance? Preserving the Past in a Digital Era,” American Historical Review 108, no. 3 (2003): 735-762; Tim Hitchcock, “Confronting the Digital: Or How Academic History Writing Lost the Plot,” Cultural and Social History 10, no. 1 (2013): 9-23; Ludmilla Jordanova, “Historical Vision in a Digital Age,” Cultural and Social History 11, no. 3 (2014): 343-348; Lara Putnam, “The Transnational and the Text-Searchable: Digitized Sources and the Shadows They Cast,” American Historical Review 121, no. 2 (2016): 377-402. Toni Weller, ed. History in the Digital Age (New York: Routledge, 2013). 4 Vincent Larivière, Stefanie Haustein, and Phillippe Mongeon, “The Oligopoly of Academic Publishers in the Digital Era,” PLOS One 10, no. 6 (2015): 5.
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1
Peeringdownthememoryhole:history,censorshipandthedigitalturnGlennTiffert1
Economic,technologicalandpoliticalforcesarefundamentallyreshapinghowknowledgeisproducedandconsumed.Muchasclaytabletsandbonegavewaytopapyrusandparchment,paperisnowyieldingtobitstreams.Inturn,librariesarerebalancingtheircollectionsawayfromphysicalvolumestowardsonlineresources,whereasmallnumberofremoteprovidersdominatestorageanddelivery,andeffectivelygovernthetermsofaccessforall.Formany,printedbooksandperiodicalsarefastbecomingremindersofthingspast.2Thisdigitalturnisseductive;itpromisestodeliverabundantknowledgewithunprecedentedconvenienceandimmediacy.Yet,commentatorshavenotedthatitalsoconcentratespower,exacerbatesinequalities,andraisesthornytheoretical,ethical,andmethodologicalconcerns.3Forinstance,ahandfulofmediaconglomeratesnowcontrolmorethanhalfofallacademicpublishing,whichmakesithighlyvulnerabletorent-seeking,poorstewardship,ordeliberatemanipulation.Theoligopolisticpricingpowerpublishersenjoy,protectedbycopyright,deliversenviableprofitmarginswhilesqueezingeventherichestinstitutions.4Thoughthequestionsaremany,censorshiphasscarcelyregisteredamongthem.Arguably,thisreflectstheprivilegedpositionofthosewhodominatetheconversation:scholarswhostudyandworkincomparativelyliberalsocieties,whereonecanreasonablyassumethatapublicationaccessibletodaywillremainsotomorrow,allowingfornormalconservation.Thisstudydispelsthatinnocence.UsingacasefromthePeople’sRepublicofChina(PRC),itexploressomeofthemostdisturbingramificationsofourdeepeningdigitaldependence.Simplyput,theChinesegovernmentisleveragingtechnologytoquietlyexportitsdomestic
1GlennTiffertistheLieberthal-RogelCenterforChineseStudiesDistinguishedPostdoctoralFellowinResidenceattheUniversityofMichigan,AnnArbor.TheLieberthal-RogelCenterforChineseStudiesgenerouslyfundedthisstudy.AttheUniversityofMichigan,ProfessorsMaryGallagher,StevenAbney,andKerbySheddenofferedvaluablefeedback.FuLiangyu,theuniversity’sChineseStudieslibrarian,helpedtolocateandacquirematerials.LuoFusheng,ArdenShapiro,YanWei,andYinShiyuanprovidedessentialresearchassistance.MargaretOrtoncodedthenaïveBayesclassifier.2Watanabe,Teresa.“UniversitiesRedesignLibrariesforthe21stCentury:FewerBooks,MoreSpace.”LosAngelesTimes(2017):AccessedJuly26,2017.http://www.latimes.com/local/lanow/la-me-college-libraries-20170419-story.html.3RoyRosenzweig,“ScarcityorAbundance?PreservingthePastinaDigitalEra,”AmericanHistoricalReview108,no.3(2003):735-762;TimHitchcock,“ConfrontingtheDigital:OrHowAcademicHistoryWritingLostthePlot,”CulturalandSocialHistory10,no.1(2013):9-23;LudmillaJordanova,“HistoricalVisioninaDigitalAge,”CulturalandSocialHistory11,no.3(2014):343-348;LaraPutnam,“TheTransnationalandtheText-Searchable:DigitizedSourcesandtheShadowsTheyCast,”AmericanHistoricalReview121,no.2(2016):377-402.ToniWeller,ed.HistoryintheDigitalAge(NewYork:Routledge,2013).4VincentLarivière,StefanieHaustein,andPhillippeMongeon,“TheOligopolyofAcademicPublishersintheDigitalEra,”PLOSOne10,no.6(2015):5.
2017
2
censorshipregimeabroadand,bymanipulatinghowobserverseverywherecomprehenditspast,present,andfuture,itisenlistingthemwithouttheirconsentinanalarmingprojecttosanitizethehistoricalrecordandglobalizeitsowncompetingnarratives.5Thepracticesexposedhereinstantiateaproteanhazardthatconfrontsusall.Butinthepagesthatfollow,Iaimtoillustratetheirgravityinawaythatalsoshowcaseshowtechnologycanenhanceourcapacitytoresist.Thestudythereforeconcludeswithsomereflectionsonwhatistobedone,followedbyanappendixdetailingthetoolsIused,myworkflow,andmethodology.Surely,onegooddigitalturndeservesanother.DefiningtheProblem:Forconsumersandproducersofknowledge,censorshipisafirstorderproblem.Itisepistemologicallycorrosivebecauseitcontaminateseverythingittouches,andevensoundresearchpracticemayoffernodefense.Perversely,themorefaithfulscholarsaretotheircensoredsources,thebettertheymayunwittinglypromotethebiasesandagendasofthecensors,andlendthosetheindependentauthorityoftheirprofessionalreputations.Historiansaretaughttomitigatesuchrisksbyreadingtheverisimilitudeoftheirsourcescritically,butthistrainingtendstotreatsourcesasastablebequestandframecensorshipretrospectivelyratherthanasadynamicgoingconcern.6Thatmustchange.Inthepast,censorsalteredhistorybystrikingoffensivepassages,tearingoutpages,andseizingordestroyingentiretexts,allcrudemethodsbytoday’sstandards.Now,theycantinkerendlesslywiththedigitalrecordtoachievetheirgoalswithouteverleavingtheirdesks,makingonenon-destructiveeditafteranother,eachpropagatingnearlyinstantaneouslyaroundtheglobe,leavingbehindnodiscernibletraceorlooseends.Algorithmscanautomatetheprocessandcatapultitfarbeyondhumanlimitations.Inshort,technologyhaselevatedOrwell’sgrotesquemetaphorofhistoryasapalimpsest,“scrapedcleanandreinscribedexactlyasoftenaswasnecessary”toaterrifyinglevelofefficiency.7InthePRC,thisisalreadyhappening.TheChinesegovernmentfusesthehegemonicinstinctsandmobilizationalinfrastructureofaclassicLeninistregimewithdeeppocketsandcutting-edgetechnologies.Itsaggressivepolicingofsocialmedia,art,news,andscholarshipsetsthebarforlike-mindedgovernmentseverywhere,earningitthesobriquet“ThePeople’sRepublicofAmnesia.”8Lately,Chineseleadersatthehighestlevelshavesingledout“historicalnihilism,”
5HeYitingZÁÆ,“ErshiyishijishiZhongguohuayufuxingdeshiji.”;K�=����'ir��=�[The
Twenty-FirstCenturyistheCenturyoftheRevivalofChineseDiscourse]Xuexishibao h�d[StudyTimes](May29,2017):AccessedJune24,2017.http://www.studytimes.cn/zydx/GCFT/2017-05-29/9532.html.6ArletteFarge,LeGoûtDeL’Archive(Paris:Seuil,1989).7GeorgeOrwell,NineteenEighty-Four(NewYork:AlfredA.Knopf,2009),42.8LouisaLim,ThePeople’sRepublicofAmnesia:TiananmenRevisited(NewYork:OxfordUniversityPress,2015).GaryKing,JenniferPan,andMargaretE.Roberts,“HowCensorshipinChinaAllowsGovernmentCriticismButSilencesCollectiveExpression,”AmericanPoliticalScienceReview107,no.2(2013):1-18;GaryKing,JenniferPan,
3
shorthandforchallengingorthodoxnarrativesaboutthepast,forspecialattention.Thosewhorunafoulofthecensorsconfrontaslidingscaleofpenalties,includingharassmentbytheauthorities,closureofpublicationsandonlineaccounts,humiliatinginvestigationsintopersonalaffairs,businessactivitiesandtaxstatus,andultimatelyunemployment,eviction,andcriminalprosecution.Arecentprovisionoflawalsoexposes“thosewhoinfringeuponthename,likeness,reputation,orhonorofahero,martyr,andsoforth,harmingthesocietalpublicinterest”tocivilliability.9Facingsuchpressure,previouslyoutspokenintellectualsandactivistsaregoingsilent.Howdoesthisimpactus?Ordinarily,scholarsexpectthatwhenonlineplatformssuchasJSTORorProQuestcarryapublication,thoseplatformswillfaithfullyreproducetheoriginalandnoteanymaterialomissionsorchanges.Thisfreeslibrariestoclearredundantvolumesfromtheirstacksorforegopurchasingthosevolumesaltogether,whichcantrimexpensesandopenupvaluablespaceforotherpurposes.Recently,anumberofsuchplatformshaveappearedinthePRCand,disarmedbyfamiliar-lookingveneers,foreignlibrariesareextendingthemcomparabletrust.Thisisdeeplymisguided.Chineseplatformsmustcomplywiththedemandsofamercurialdomesticcensorshipregimeandtheauthoritariangovernmentbehindit.Inpurelycommercialterms,thisprecludesthemfromdeliveringfullreplacementvaluetotheirsubscribers,orguaranteeingthatatextaccessibletodaywillremainsotomorrow,withoutmodification.10Yet,temptedbytherichesonoffer,librariesareincreasinglyoutsourcingcriticalpartsoftheirChinesecollectionstotheseplatforms.Motivatedbythriftandefficiency,somearegoingstillfurtherbyunilaterallydeprecatingtheirphysicalholdings,analarmingdevelopmentthatnotonlycedescontrolofacrucialportionofthesourcebasefromwhichtomorrow’sscholarshiponChinawillbewritten,butalsoeliminatestheindependentevidenceessentialfordetecting,exposing,andcircumventingtheexerciseofcensorshiptodayandinthefuture.Letusmakethisconcrete.WhenthePRCwasestablishedin1949,thenewcommunistgovernmentabrogatedtheinheritedlegalsystem,andproposedtoconstructasocialist
andMargaretE.Roberts,“HowtheChineseGovernmentFabricatesSocialMediaPostsforStrategicDistraction,NotEngagedArgument,”AmericanPoliticalScienceReview(forthcoming).SarahCook,TheLongShadowofChineseCensorship:HowtheCommunistParty’sMediaRestrictionsAffectNewsOutletsAroundtheWorld(WashingtonDC:CenterforInternationalMediaAssistance,NationalEndowmentforDemocracy,2013).9JoshChin,“ChinaSettoTweakCivilCodetoPunishRevisionsofMartyrLore.”WallStreetJournal:ChinaRealTimeReport(2017):accessedJune17,2017,http://on.wsj.com/2miOVcF.ChiFangxu¨&Ê,“‘Minfazongze’di
185tiaodehexinyaoyishiweihushehuigonggongliyi.”ÏbMG�Ð+ 185R�����c�� :zf£[SafeguardingtheCommonInterestsofSocietyistheCorePointofArticle185ofthe'GeneralPrinciplesofCivil
Law']QiushiX�[SeekingTruth](June7,2017):accessedJune17,2017,http://www.qstheory.cn/dukan/hqwg/2017-06/07/c_1121099778.htm.10ForeignsubscriberstotheonlineeditionofthePeople’sDaily�b�d,forexample,maygetdifferentresultsdependingoniftheirvendorhostsitsserversinChinaornot.Somesearchtermsmaynotgenerateanyresultsandinsteadtriggeranautomaticconnectionresetthatlockstheuseroutofthedatabaseforaminuteormore.
4
successorfreeofcontaminationfromtheancienrégime.Intheyearsthatfollowed,thePRC’stwoleadingacademiclawjournalsrecordedthesometimesfiercestrugglesoverwhatthisnewlegalsystemshouldlooklike.Zhengfayanjiu�"&'[Political-LegalResearch]comprised62issuesbetween1954and1966,whentheCulturalRevolutionforceditsclosure.SponsoredbytheChineseAssociationforPoliticsandLaw,inBeijing,itcountedmanyofthehighestlegalofficialsinthecentralgovernmentamongitspatrons,anditscoveragegenerallyfavoredtheirstatistpriorities.Bycontrast,Faxue"�[LawScience]enjoyedamuchshorterlife.Itcomprisedjusteighteenissues,allpublishedbetween1956and1958,whentheAnti-RightistCampaignforceditsclosure.11FaxuewassponsoredbytheEastChinaInstituteofPoliticsandLawinShanghai,oneofahandfulofregionalacademiesestablishedintheearly1950sbythePRCstatetotrainanewgenerationofsocialistcadresforadministrativeandlegalpositions.NoinstitutionintheUnitedStatesholdsacomplete,originalprintrunofZhengfayanjiu,andFaxueisnearlyasrare.Somehavemicrofilmorreprinteditions,usuallystoredoff-site.12Formostresearchers,onlineaccessisthenorm.TwoofthefourleadingonlineplatformsinthePRCofferfull-textarticlesfromthesejournals:ChinaNationalKnowledgeInfrastructure��%)[CNKI],acommercialventureconnectedtoTsinghuaUniversity,andtheNationalSocialSciencesDatabase��������[NSSD],agovernment-sponsored,open-accessplatform.13Mindfulofthat,thisauthorcleareddozensofarticlesphotocopiedlongagofromthosejournalsoutofhisfiles,butnotwithoutfirstconfirmingtheavailabilityofeachasadownload.Curiously,someofthearticlesweremissingonline,andtheomissionssuggestedapattern.Aresearchprojectwasborn.Myawarenessthattheseomissionswereworthyofinvestigationsprangfrommygroundinginthesubjectmatter.Ihavereadwidelyintheseparticularjournalsandotherslikethem,conductedyearsoffieldworkinChinaontheearlyPRClegalsystem,includinginstatearchives,andhaveintervieweddozensoflegalcadreswhowereactiveinthatperiod,someofwhom
11Thefirstthreeissues,datingfrom1956,bearthetitleHuadongzhengfaxuebaogH�M d[EastChinaJournalofPoliticsandLaw].ThepublicationwasrenamedFaxuebeginningwiththefirstissuein1957.12Afive-volumereproductionofZhengfayanjiuwaspublishedinTokyoin1967.TheissuesfoundintheUnitedStatestendtobefromthisedition.Zhengfayanjiu(Tokyo:Da’an,1967).Atwo-volumereprintoftheFaxueissuesexaminedinthisstudywaspublishedinShanghaiin2012.Owingtocompetitionfromtheonlineedition,veryfewinstitutionsintheUnitedStatesownit.WhenIattemptedtobuycopies,PRCvendorsinformedmewithoutexplanationthatthevolumescouldnotbeshippedtomyoverseasaddress,thoughotherbooksonChineselawinthesameorderspresentednoobstacle.Iobtainedthemthroughotherchannelsanddiscoveredthatthereprintsareverygoodbutnotperfectreplacementsfortheoriginals;virtuallyallofthecontenthasbeenaccuratelytranscribed,includingthearticlesmissingfromtheonlineedition,buttheoriginallayouthasnotbeenpreservedandsomeofthesidebarsaremissing.HeQinhuaZ°g,ed.Faxuedelishi:1956nian-1957nianM �~�:
1956�-1957�[TheHistoryofFaxue:1956-1957](Beijing:Falüchubanshe,2012);WangLimin�ub,ed.Faxue
delishi:1958nianM �~�:1958�[TheHistoryofFaxue:1958](Beijing:Falüchubanshe,2012).13Theothertwovendorseitherlackcoverage(WanfangDigitalKnowledgeServicePlatformk&]�"yUwTl),ormerelyindexthearticleswithoutofferingfull-textdownloads(Duxium¤).
5
wroteforZhengfayanjiuandFaxueandparticipatedfirsthandinthecontroversiestheydocumented.Theinterpretationofthedatapresentedbelowrestsonthatfoundation.Thefirststepintheprojectwastogaugethescaleoftheomissionsbycomparingtheonlineholdingsagainsttheoriginalpaperissues,andrecordinganydiscrepancies.14Thus,IpiecedtogetheracompletesetoforiginalpapereditionsofbothjournalsdrawnfromthecollectionsofseverallibrariesaroundtheUnitedStates,supplementedbytargetedacquisitionsfromChina.Toensurecommensurabilityandkeepthescopeoftheprojectmanageable,Idecidedtoconfinemyanalysistoarticlespublishedfrom1956throughtheendof1958,theonlyperiodwhenbothjournalsoverlapped.Fortunately,thisalsopermitsonetojuxtaposetheHundredFlowersCampaignagainsttheAnti-Rightistbacklashthatabruptlyfollowed,whentheCCP’sbriefsolicitationofearnestfeedbackswitchedintosearingretribution.Fortheyearsinquestion,bothCNKIandNSSDhaveuniformholdings,anduniformgapsincoverage.Thatistosay,theyarebothmissingexactlythesame30articlesfromZhengfayanjiu,andexactlythesame33articlesfromFaxue,whichsuggestsatleastthreepossibleexplanations.First,thevendorsmayhavebeenworkingfromthesameincompletesetofsourcematerial,orfromdifferentsetswithcoincidentallyidenticalgaps.However,everyissueinmysamplereceivesatleastsomecoverageonline,whichimpliesthatthemissingarticleswereathandwhentheissueswereoriginallyscanned,andthereforedeliberatelyomitted.Furthermore,thenotionthatbothvendorswouldhaveidenticalgapspurelybyhappenstancestrainscredulity.Thepersistenceofthatuniformityalsodefiessoundbusinesssense.Afterall,onewouldexpectacompetentvendortocorrectinadvertentomissions.Neitheronehas.Tofollowup,IusedanonlinecustomerserviceformtoreportseveralmissingarticlestoNSSD.Withinacoupleofweeks,asuccessionofformulaicresponsesarrivedbyemail:“We’resorry.Aftermanysearches,westillcannotfind:‘(articletitle).’Thanksforyourinquiry!”Thesecondpossibilityisthatonevendorissimplymirroringthecontentoftheother.Atfirstglance,thisappearsplausible,butthemirroringwouldhavetobelimitedinscopebecausethetwovendorsdifferintheircoverageofotherjournalsfromthatperiod.Finally,thesimplestexplanationconsistentwithalloftheobservableevidenceisthatbothvendorsaresubjecttoacommonaffirmativerestraint,perhapsanauthoritativeblacklistorcensorshipdirectiveimposedfromabove.Thisconclusionissupportedbythetemporaldistributionofthemissingarticles.Figures1and2showthatthegapscommenceabruptlyinlate1957andthentaperoffas1958progresses,correspondingexactlytothefirstarcoftheAnti-RightistCampaign.TheAnti-RightistCampaignunleashedaferociouswaveofpoliticaldenunciationsandpersecutionsacrossChinathatsweptupasmanyas700,000people,andusheredinwhatChinesejuristshavesincecalledaperiodof
14ThecomparisonsweremadeinNovember2016,andagaininFebruary2017.
6
“legalnihilism.”15ThecampaignremainsahighlysensitivetopicinthePRC,andstatecensorshipstillrestrictsinquiriesintoit.
Figure1:PagesMissingbyIssue,Zhengfayanjiu(1956-58)
Figure2:PagesMissingbyIssue,Faxue(1956-58)
Someoftheaffectedissuesaremissingaroundhalfoftheirtotalpagecount,andthegapsareconcentratedalmostexclusivelyatthetopsoftheirtablesofcontents.Ineffect,theseissueshavebeendecapitated,shornoftheirleadarticles.ConsiderthefourthissueofFaxuefrom1957,whichfirstappearedinAugustofthatyear.ThetableofcontentsinFigure3was
15ZengXianyitÉ�,“XinZhongguofazhi50nianlunlüe,”���M�50�a�[OnRulebyLawinthePast50
YearsSincetheFoundingofNewChina]Zhongguorenmindaxuexuebao���b� d[JournalofRenminUniversityofChina]6(1999):31-41.Ahandfulofscatteredarticlesfromearlierissuesofthesejournalshavealsobeenomittedonline,forexample,a1956FaxuearticlebytheSovietjuristS.N.Bratus.C.H.BolatuxiÂvLD,
“GuanyuSuweiaiyibanfaxuelilunkexuegongzuozhongdejigewenti”?0 c¾��M <a� O2��5�CP[SeveralQuestionsintheScientificWorkonGeneralLegalTheoryintheSovietUnion],FaxueM [LawScience]3(1956):1-4.Analyzingtheiromissionisbeyondthescopeofthisessay.
0%
25%
50%
75%
100%
Year.Issue
Present
Missing
0%
25%
50%
75%
100%
Year.Issue
Present
Missing
7
scannedfromanoriginalpaperedition.Redarrowsmarkthearticlesthataremissingfrombothonlineplatforms.Theyaccountfor30oftheissue’s72pages.Themissingtitlesappearintranslationalongside.
Figure3:TableofContents,OriginalPrintEditionofFaxue1957(4)
ArticlesMissingOnline
1. RepulsetheRightistAttackonthePeople’sLegalSystem
2. Refute“TheLegalProfession(isSplit)BetweenPartyandNon-Party”
3. TheDictatorshipoftheProletariatisaRegimethatDoesNotAcceptAnyLegalLimitationsSidebar:TheRelationshipbetweenYangZhaolongand
RoscoePound
4. OntheNecessityofDictatorshipandtheLegitimacyoftheCampaigntoEliminateCounterrevolutionaries
5. WeMustThoroughlyAbrogatetheSham(Nationalist)SixCodesSidebar:ExcavatingYangZhaolong’sFascistRoots
6. American“RuleofLaw”TravelstheRoadofFascistDictatorship
7. TheShanghaiLegalProfessionExposesandCriticizestheAnti-PartyWordsandDeedsofRightistElementsWangZaoshiandYangZhaolong
8. TheTrueFaceofHypocriteWangZaoshi
9. YangZhaolong:LoyalMinistertotheChiangFamily,DiscipleofRoscoePound
Manyresearchersinteractwithonlinedatabasesprimarilythroughtheirsearchengines.BothChineseplatformsofferaconventionalsuiteofsearchfunctionstohelpuserslocatecontent,buttheirsearchenginesareofcourseblindtothemissingcontent,returningonlysanitizedresultsandleavingtheendusernonethewiser.Byquietlyomittingcertainvoicesandopinions,theyaredistortingoursenseofwhatthediscourseonagiventopicactuallylookedlike,wheretheweightofopiniononitmayhavebeen,howthatmighthavechangedovertime,andwhy.Notonlyaretheycomplicitintheintentionalmisrepresentationofhistory,buttheyarealsocontaminatingresearchbasedonthatsourcebaseandviolatingthetrustoftheirusers.Perhapsthisiswhytheyconcealthecull.Userswhohappentodrilldowntotheonlinetableofcontentsforanaffectedissuewillfindanunbrokenlistofarticleswithnoplaceholdersornotationsformissingcontent,andmayconsequentlyneverrealizethatanythingisamiss.
8
(Figure4)Ononesite,theonlyhintwouldbeunexplainedgapsinthepagenumbersequenceofthearticles,adetaileasytooverlookormissthesignificanceof.Theothersiteomitseventhiscluebyforegoingpagenumbersaltogether.
Figure4:OnlineTablesofContentsforFaxue1957(4)(accessedJune7,2017)
ChinaNationalKnowledgeInfrastructure(CNKI)
NationalSocialSciencesDatabase(NSSD)
9
Analysis:Havingpinpointedtheanomalies,letusnowprobetheirsubstancewithamixtureofcomputationalandqualitativeanalysis.Instatisticalterms,languagetendstobenoisyandweaklydifferentiated,particularlywhenthedocumentsinone’scorporaallbelongtothesamegenre,period,andsubjectmatter,suchasChineselawreviewarticlesfromthelate1950s.TheconstraintsonfreeexpressioncommontoLeninistregimescanmagnifythishomogeneity.Myfindingsareneverthelessedifying.Mycorporaconsistofeveryarticleandsidebarpublishedinbothjournalsfrom1956through1958,adatasetcomprising737documentsandnearlyfourmillioncharacters.Ibeganbyaskingseveralbasicquestionsaboutthem.Aretheonlineomissionstopicallystochasticordeterministic?Ifstochastic,Iwouldexpecttoseearandomdistributionoftopicsacrossthesubsetsofmissingandincludedarticles,withnoobviousfeaturesseparatingonefromtheother.Thiswouldmakethecaseforintentionalcensorshiphardertosustainunlessthecriteriaareessentiallyarbitrary,orperhapsexogenoustomyvariablesandbeyondwhattheyareabletocapture.Ifdeterministic,Iwouldexpecttoseestructureinthedata,somedegreeofclusteringorlumpinesswithinthemissingandincludedsubsets.AssumingIdetectedstructure,thenextstepwouldbetocharacterizeit,toaskwhichfeaturesaccountfortheomissionofsomearticlesandnotothers,andtomeasurethestrengthoftheireffects.Strong,unambiguousfindingswouldnotonlysupportthecasefordeliberatecensorship,butalsoeffectivelyreverseengineerthecriteriathattriggerit,andhintatthemotivationsbehindit.ThescopeofmycorporaandthewayIcodedthemallowedmetoexplorethesequestionsacrossbothtimeandspace(Beijingvs.Shanghai).InFigure5,thespatialarrangementofthedotsexpressestherelativeproximityofmytextstooneanotherbasedontheirdiscursive(cosine)similarity.16Eachbluedotrepresentsadocumentintheonlinecorpus,andeachreddotrepresentsadocumentmissingfromit.Thereddotsalsocarrylabelstoidentifythenamesofthespecificdocumentstheyreferto.OnecanplainlyseethattheFaxuecorpusexhibitssharperinternaldifferentiationthantheZhengfayanjiucorpus,andthisresultholdsacrossmyotherstatisticaltests.Apartfromafewstraymembers,nearlyalloftheFaxuearticlesmissingonlineinhabittheupperrightquadrantofthesphere,whichistosaythatinspiteofvariationintheirauthorship,topics,andwording,asagrouptheydemonstrateanunmistakablecoherence.ThisfindingstronglysuggeststhattheomissionoftextsfromtheonlineeditionofFaxueisnottopicallyrandom,butratherinvolvesadiscriminatinglogic,thoughwecannotyetsaywhatthatis.
16Forfurtherinformation,seetheappendix.
10
Figure5:CosineSimilarityMeasurements
Zhengfayanjiu Faxue
WecanuncoverhiddendifferentiationintheZhengfayanjiucorpusbyusingadifferentstatisticalprocedureandaddingathirddimension.(Figure6)Nearlyallofthemissingdocumentsfalloutsideofthedensebluecore,whichindicatesthatsomethingaboutthem,too,isdifferent,thoughnottotheextentthattheycleavecleanlyaway.Recall,theystillbelongtoacommongenreandfieldofknowledge.
Figure6:PrincipalComponentAnalysis(3D)
Zhengfayanjiu Faxue
Havingestablishedthattheomissionsareunlikelytoberandom,wecannowtrytoanalyzetheircharacteristics.Aprocedurecalledsphericalk-meansclusteringreadsacorpusandcomputesauser-definednumberofclustersthatthealgorithmthenpopulateswithdocumentsarrangedspatiallyaccordingtotheirrelativesimilarity.Myimplementationalsocolorcodestheclusters,andreportsbackboththetermsthealgorithmselected(fromthecorpus)tobuild
11
eachcluster,andthetitlesofthearticlesbelongingtothoseclusters.Aswiththeothercalculations,thisprocedureisblindtothepresenceorabsenceofthedocumentsonline,andmerelylabelstheomitteddocumentsattheendasaconveniencetotheuser.17Figures7and8displaytheresults,includingthetermsthatdefinetheclusterswherethemostomitteddocumentsarefound.Asbefore,thedispersionofomittedarticlesbelongingtotheZhengfayanjiucorpussuggestslessinternaldiscursivesimilarity.FourteenofthethirtymissingarticlesbelongtoCluster3,definedbyabasketoftermsdescribingtheideologicaldimensionsofthelegalsystem’srelationshiptothestateandsociety.AnadditionalninearticlesbelongtoCluster4,whichcentersoncriminallaw.(Figure7)
Figure7:Sphericalk-MeansClustering
Zhengfayanjiu
Cluster3terms:thepeople,thestate,law,class,work,society,capitalistclass,-ism,democracy,leader,
socialism,dictatorship,system,revolution,represents.Cluster4terms:tocommitacrime,themasses,element,work,thepeople,policy,law,struggle,penalty,
dictatorship,counterrevolutionary,enemy,production,class,political-legal.FortheFaxuecorpus,theclusteringismuchtighter,indicatingthattheomitteddocumentsareinternallyquitesimilartooneanother.(Figure8)Moreover,mostoftheomitteddocumentsfallwithinasinglecluster.ThetermsdefiningthisclusteroverlapsignificantlywiththosefromtheprecedingZhengfayanjiucorpus,withthenotableadditionofYangZhaolong,aprominentShanghaijuristaboutwhomIwillhavemoretosaybelow.Thisconsistencyacrosscorporabeginstodispelthecloudoverpreciselywhysomedocumentshavebeenomittedonline,andcanguideatripbacktothepertinenttextsforacloser,traditionalreading.
17MyimplementationofthisclusteringalgorithmadaptscodepublishedbyBrandonRose.Rose,Brandon.“DocumentClusteringWithPython.”AccessedJune24,2017.http://brandonrose.org/clustering.
12
Figure8:Sphericalk-MeansClustering
Faxue
Cluster3terms:law,class,thepeople,thestate,dictatorship,society,torule,democracy,capitalistclass,old
(ancienrégime)law,policy,proletarianclass,YangZhaolong,-ism,socialism.Letusdivedeeperstill.Aprocedurecalled χ2(chi-squared)featureselectionmeasuresthestrengthofthedependencebetweentheappearanceofaterminatextandthemembershipofthattextintheomittedclass.Thehigherthescorethestrongerthecorrelation.Theresultsarehighlyinformativeandallowus,broadlyspeaking,toreverseengineerthelogicbehindtheomissions.Figures9and10listtheterms(features)ineachcorpusmostcloselycorrelatedwiththearticlesmissingonline,therelativestrengthofthosecorrelations,andtheirrespectivedegreesofstatisticalsignificance.Bothlistsdescribesimilarcurveswithsimilarscales.ThesoleexceptionrelatestoYangZhaolong��.,whoseextraordinaryscoreontheFaxuelistskewsitsgraph.Bothcurvesdescendquicklyintolongtails,whichindicatesthatomissionismosttightlyassociatedwithjustafewfeatures.
13
Figure9:χ2FeatureSelection(α=.05)
Zhengfayanjiu(χ2score,pvalue)
Figure10:χ2FeatureSelection(α=.05)
Faxue(χ2score,pvalue)
Onthisevidence,Iamcomfortableattributingtheomissionstodeliberate,targetedcensorship.Notethattheselectedfeaturesdiffersignificantlyfromonejournaltotheother,whichimpliesthatthecensorsarerespondingtodifferentproximatetriggers.Onecanseethisplainlywithrespecttopeople.PersonalnamesaccountformorethanhalfofthefeaturesontheZhengfayanjiulist.TheseincludeprominentacademicswithrootsintheformerNationalist
7.52,p=.0066.84,p=.009
6.09,p=.0145.33,p=.0215.24,p=.0225.23,p=.022
4.79,p=.029
4.58,p=.032 4.27,p=.039 4.19,p=.041
3.95,p=.047
0 1 2 3 4 5 6 7 8 9 10
RightistElementWangHan
RightistWangJixinLuMingjian
CampaigntoEliminateCounterrevolutionariesToweightheevidence
DeduceQianDuansheng
JiaQianWuChuanyi
45.53, p=1.50E-11
10.11,p=.0016.35,p=.0125.84,p=.016 5.62,p=.018 5.05,p=.025 4.95,p=.026 4.90,p=.027 4.84,p=.028 4.82,p=.028 4.76,p=.0294.73,p=.030 4.64,p=.031 4.48,p=.034 4.48,p=.034 4.34,p=.037 4.04,p=.0453.87,p=.049
0 5 10 15 20 25 30 35 40 45 50
YangZhaolongWangZaoshiRuleoflawKuomintang
CampaigntoEliminateCounterrevolutionariesRightistElementRoscoePound
NewsSchools&Departments
ScientificLaw
RuleofmanLuoJiaheng
ChinaDemocraticLeagueRightist
LegislationFascist
(Nationalist)SixCodes
14
establishment,suchasQianDuansheng+(andWuChuanyi��,,aswellasveteranPartycadresandhighofficials,suchasLuMingjian-��andJiaQian*$oftheSupremePeople’sCourt.Bycontrast,theproportionofnamesontheFaxuelistisfarsmaller,andnoneoftheidentifiedindividualswerePartymembersorhighofficials.TheFaxuelistmakesupthedifferencewithfeaturessuchas“Kuomintang,”“legislation”and“ruleoflaw.”Furthermore,thenamedindividualsaregeographicallysegregated,areflectionofhowtiedeachjournalwastotheconcernsofitsrespectivepublisherandsponsors.AllofthepeopleontheZhengfayanjiulistwerelocaltoBeijing,andallbutoneontheFaxuelistwerelocaltoShanghai.ThesoleexceptionwasRoscoePound,theformerdeanofHarvardLawSchool,whomadethelistbyvirtueofhisclosepastassociationwiththetopfeature:YangZhaolong.Onecouldsurmisefromthisthatthecensorsarepursuingdifferentaimsfromonejournaltothenextbut,asthek-meansclusterscounter-indicate,thecontendingfeaturelistsactuallyshareacommonagenda:buryingtheevidenceeachjournalrespectivelybetraysofthecontroversiesthatrenttheemergentPRClegalsystem.Inmid-1957,buddingdebatesovermatterssuchasjudicialindependence,thetranscendenceoflawoverpoliticsandclass,thepresumptionofinnocence,andtheheritabilityoflawabruptlygavewaytovituperativedenunciationsofthoseideasandtheirsympathizers.EverypersonlistedinFigures9and10(savePound,ofcourse)wascaughtinthatshiftandpersecutedasanexampletoothers.Themajorityoftheremainingselectedfeaturesalsorefertothatbacklash.Bysanitizingtherecord,thecensorshavedistortedourunderstandingofthishistory,andofthestrugglestoshapetheyoungPRClegalsystem.ThisconvenientlyabsolvestheCCPfromhavingtofaceuptothemortifyingcontemptforlawthatdefineditsearlyhistory,anditssuppressionofthosewhodissentedfromthat.Toappreciatehowthecensorsarepurposefullyrewritinghistory,wecanexaminethetoptermbykeynessfromeachjournal–“rightistelement”�#��(Zhengfayanjiu)and“YangZhaolong”��.(Faxue)–aswellasatermwithalowerkeynessscorethatisnonethelessofgreatsignificancetoday:“theruleoflaw”"!.18Figures11-14plottheweightofallthreetermsovertime,arrangedinthesequentialorderofthearticlesinwhichtheyappear,andcolor-codedaccordingtothecensorshipstatusofthosearticles.19Thiscastsaspotlightonexactlywhichargumentsthecensorsareselectingforandagainst,andthediscursiveeffectthatsortinghas.
18Isortedthetf-idfmatricesofmycorporaaccordingtothe“keyness”oftheir30,000+uniqueterms,heredefinedastheabsolutedifferencebetweenthemeantf-idfscoresforatermacrossallofthearticlesinthecensoredclass,andallofthoseintheuncensoredclass.Thisprovidesarankingofhowlopsidedtherepresentationofakeytermisbetweenthetwoclasses,andthereforehelpsustograsptheuniqueinformationthatisbeinglostwhencertainarticlesareculled.Thetop10termsbykeynessforeachjournalwereasfollows.Zhengfayanjiu:rightistelement,rightist,class,thepeople,legalsystem,leadership,struggle,element,capitalistclass,judicial.Faxue:YangZhaolong,Law,class,WangZaoshi,dictatorship,capitalistclass,thepeople,Kuomintang,rightistelement,democracy.19Weightreferstoaterm’stf-idfscoreinagivendocument.
15
Figure11:RightistElement,Zhengfayanjiu(1956-58)
TakingFigure11asanexample,wecanseethattheterm“rightistelement”burstintousagequitesuddenlyatthetemporalmidpointofmythree-yearsample,whichcoincidesexactlywiththeoutbreakoftheAnti-RightistCampaigninmid-1957.Afterashortperiodofhighintensity,whenrightistswerehounded,usagedroppedoffsignificantlyasthecampaignranitscourse.Thegraphshowsthatusageoftheterm“rightistelement”isroughlysplitbetweentheuncensoredandcensoredclassesinthisjournal,andthetermappearswithsimilarweightacrossboth.Unpackingthatfurtherisbeyondthegoalsofthisessay,butthestrengthofthetechniqueisthatonecaneasilytraceeachdatapointbacktoitscorrespondingarticleforcloserinspection.Forpresentpurposes,wecansimplysaythatabouthalfofthe“rightistelements”intheAnti-RightistCampaignhavegonemissingfromtheonlineversionofthisjournal.ForanyoneinterestedinthehistoryofthePRC,thatshouldsetoffalarmbells.Itcallsthereliabilityandgoodfaithoftheonlineplatformsintoseriousquestion.
Figure12:YangZhaolong,Faxue(1956-58)
0
0.05
0.1
0.15
0.2
0.25
0.3
tf-id
fscore
Year.Issue
Uncensored
Censored
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
tf-id
fscore
Year.Issue
Uncensored
Censored
16
ThewarpingimpactofthecensorsisevenmorepronouncedforthetoptermintheFaxuecorpus:YangZhaolong(1904-1979).(Figure12)Yangwasoneofthemostoutstandingandinternationally-respectedChinesejuristsofhisgeneration.AgraduateofHarvardLawSchool(SJD’35),heheldavarietyofhighacademicandgovernmentalpostsintheNationalistera,includingchiefoftheMinistryofJustice’sCriminalDivision,wherehedirectedChineseparticipationintheTokyowarcrimestrials,andactingProcurator-General.Inspiteofthesepositions,Yang,likemanycommittedlegalreformers,grewdeeplydisenchantedwiththeNationalistgovernment.In1949,undergroundCCPoperativespersuadedhimtoremainonthemainland.Inshortorder,thePRCgovernmentappointedhimDeanofSoochowUniversityLawSchool,hisundergraduatealmamater,butwhenitclosedtheschoolin1952,hejoinedthemajorityoftheformerNationalist-eralegalcommunityinprofessionalexile,barredfromplyinghisvocation.Duringabriefthawfrom1956toearly1957,YangwasinvitedtojoinFudanUniversity’slawfaculty,andtheeditorialboardofFaxue.Inthosefewmonths,hecontributedarticlestoFaxuethatcogentlyrefutedCCPorthodoxyontheclassnatureandheritabilityoflaw,andoncauseandeffectincriminallaw.20HealsodrewattentioninpublicforatotheslowpaceofcodificationinthePRC,thelowqualityofPartylegalpersonnel,andofficialdiscriminationagainsthighly-trained,non-Partyexpertslikehimself.21Forhisprestigeandfrankness,Yangpaidaheavyprice.WhentheAnti-RightistCampaignstruckShanghai,awaveofcalumnyenvelopedhim.20YangZhaolong¯Ã�,“Falüdejiejixinghejichengxing”M��·^>�s�>[OntheClassNatureand
HeritabilityofLaw],HuadongzhengfaxuebaogH�M d[EastChinaJournalofPoliticsandLaw]3(1956):26-
34;YangZhaolong¯Ã�,“Xingfakexuezhongyinguoguanxidejigewenti”¼M� �6%�5�CP
[SeveralProblemsintheRelationshipbetweenCauseandEffectintheScienceofCriminalLaw],FaxueM [LawScience]1(1957):61-63.21“Peiyangxinshengliliang,haiyoubushaowenti”¥p��(N,���FCP[TrainingUpaNewForce,There
AreStillManyProblems].Xinminwanbao�b8d[NewPeople'sEveningPost],May4,1957:1-2.“Shanghai
zhishijietanguanchebaijiazhengmingwenti”�I"yJ�À«`��ºCP[Shanghai'sIntellectualCommunityDiscussestheProblemofImplementing'LetOneHundredFlowersBloom,LetOneHundredSchoolsContend'].GuangmingribaoS*�d[EnlightenmentDaily],May1,1957:2.“Sifagongzuo‘qiang’gao‘gou’shen,
Minmengshiweizhaokaisifazuotanhuipangtingji”oMO2'©'$'´'_,b¹[�»�1� ¢@Q[The'Walls'AreHighandthe'Chasms'DeepinJudicialWork,NotesfromaForumConvenedbytheMunicipalCommitteeoftheChinaDemocraticLeague].Xinminbao�bd[NewPeople'sEveningPost],May19,1957:1.
YangZhaolong¯Ã�,“FalüjiedangyufeidangzhiJian”M�J±7e±#.[TheSplitBetweenPartyandNon-
PartyintheLegalCommunity].Wenhuibao,³d[WenhuiDaily],May8,1957:2;YangZhaolong¯Ã�,
“Woguozhongyaofadianheyichichihaibubanbu?”��AM�Z�¨¨��½q?[WhyAfterSoLongHave
China'sKeyLegalCodesNotBeenPromulgated?].Xinwenribao�{�d[DailyNews],May9,1957:2-3;Yang,
Zhaolong,“Wotanjidianyijian”��5�-9[IDiscussSeveralPoints].Xinwenribao�{�d[DailyNews],June6,1957:3.
17
TheoriginalprintversionofFaxuerecordsYang’sremonstrationsandtheorchestratedcampaignagainsthim,buttheonlineeditionpresentsafardifferentview.ThecensorshaveallowedthetwoconceptualarticlesYangwroteforthejournaltostand,butnothiscritiquesofthelegalsystem’spracticaldefectsorthesavagerebuttalsheendured.Thosehavebeenlargelywipedaway,takingmostofhisfootprintinthisjournalalongwiththem.Infact,therebuttalsagainstYangaccountforthelargestcohortofarticlescensoredfromFaxue,whichiswhyhisscoresontheχ2featureselectionandkeynesstestseclipseallothertermsinmycorpora.ThisisalsowhatendowsthecensoredFaxuecorpuswithsuchatightinternaldiscursivecoherence.Evidently,thecensorswishustocomeawaywiththebreathtakingmisapprehensionthatYang’sargumentsweremainlytheoretical,andwellcountenanced.22Backintherealworld,however,theymadehimthetoptargetoftheAnti-RightistCampaigninShanghai’slegalcommunity,andledtotwelveyearsofimprisonmentasacounterrevolutionary.Onthiscount,YangZhaolongisexceptionalonlyinscale,notinkind.ThesamepatterngenerallyappliesacrossthefeatureandkeynesslistsIhavecomputed.MuchliketheretouchedofficialphotoofMao’sfuneral,inwhichgapsinthemourninglinebelietheerasureoftheGangofFour,thehistoryofthisperiodhasbeenintentionallymanipulatedtomaskelementstheauthoritieswouldnowsoonerforget.23Whyshouldcensors,andthegovernmentthatpresumablydirectsthem,bothertoburyevidenceofjurisprudentialbattlesfought60yearsago?Itcanonlybebecausethosebattlesbearonissuesofcardinalsignificancetoday.Sincethelate1990s,theCCPhascommitteditselftoconstructingthe“ruleoflaw”inChina,andArticle5ofthestateconstitutionspecificallydeclaresthePRCtobeasocialistcountryundertheruleoflaw.ButtheCCPhasanidiosyncratic,exceptionalistunderstandingofwhatthatmeans.24Ithasdissociateditselfnotonlyfromforeignvariantsoftheruleoflaw,butalsofromChineseonesthathappennottobesocialist,likethoseendorsedbyRepublicangovernmentsbefore1949,oronTaiwantoday.Thisleavesitlittlechoicebuttolookforlegitimacywithin,andthepickingsareslim.From1957to1976,Chinawasrockedbyasuccessionofdevastatingpoliticalcampaigns:theAnti-Rightist
22MeiNaihan¸�¬.“Guanyu‘fadejiejixinghejichengxing’detaolun:jieshaoShanghaifaxuehuidiercixueshu
zuotanhui”?0‘M�·^>�s�>’��a:¡µ�IM +;! �1� [TheDiscussionon‘TheClassNatureandHeritabilityofLaw’:IntroducingtheShanghaiLawSociety’sSecondAcademicForum].Faxue2(1957):28–30.ZhangJinghua���.“Guanyufalüjichengxingzhongdejigewenti”?0M�s�>��5�CP
[SeveralQuestionsConcerningtheHeritabilityofLaw].FaxueM [LawScience]5(1957):18–21.23Marsh,Bill,“FakedPhotographs:Look,andThenLookAgain.”NewYorkTimes,August22,2009:wk4.DavidKing,TheComradeVanishes:TheFalsificationofPhotographsandArtinStalin’sRussia(London:Tate,2014).24Zhonggongzhongyangzhengfaweiyuanhui�z�¶�M�W [Political-LegalCommitteeoftheCCPCentral
Committee],Shehuizhuyifazhilinianduben� )�M�<j|/[ReaderontheConceptofSocialistRuleofLaw](Beijing:Zhongguochang’anchubanshe,2009).
18
Campaign,theGreatLeapForward,theSocialistEducationCampaign,andtheCulturalRevolution.Noneoftheseoffersacrediblepedigreeforthecurrentprogramof“socialistruleoflawwithChinesecharacteristics,”andsotheearly1950smustcarrytheburdenalone.TheargumentsraisedbyYangandothers,andthesearingbacklashagainstthemsubvertthispossibilitybyexposingthegravityofthelegalsystem’scongenitaldefects,andtheferocityofthecontroversiesthatseizedit.Moredamningstill,theybearwitnesstowherethenation’sreveredfoundersactuallystood.Andinthis,theydenythelegalsystemtheoriginmythitcraves.Wecancorroboratethatconclusionbyanalyzinghowcensorshiphassculptedthe“ruleoflaw.”Upuntil1953,someseniorjudicialcadresinthePRCspokeapprovinglyoftheruleoflaw,buttherevolutionarypoliticsoftheJudicialReformCampaign(1952-53)forcedthemgenerallytoabandonthetermandmoveinsteadintothesafeideologicalharborofSoviet-style“socialistlegality.”The“ruleoflaw”cametobeusedderisively,asamarkerforthehypocrisybehindwhichcapitalistlegalsystemscloakedtheirclassdominationandexploitation.Furthermore,thevaluesassociatedwithit,suchasjudicialindependence,professionalism,impartiality,andproceduralrigorweresuppressedasheterodoxorcounterrevolutionary.Theleadingacademiclawjournalsoftheperiod,ZhengfayanjiuandFaxue,documentthisincontrovertibly,whichposesaproblemformythmakerstoday.Theirresponseshouldnotsurpriseus.Owingtotheideologicalclimate,invocationsofthe“ruleoflaw”inthesejournalsarefew.Thataloneshouldraiseaflag.Justthesame,censorshavemanipulatedthemwithcare.Frommydatasetwecanreconstructtheiroriginalfootprint.(Figure13)
Figure13:RuleofLaw,Faxue(1956-58)
GiventheimportancePRCauthoritiesnowattachtoestablishingapedigreefor“socialistruleoflawwithChinesecharacteristics,”onemightexpectthemtoshowcasethearticleinwhich“theruleoflaw”appearswiththeheaviestweight.Ittowersabovetheothers.Yet,theoppositeistrue;ithasbeencensored.Why?
0
0.1
0.2
0.3
0.4
0.5
tf-id
fscore
Year.Issue
Uncensored
Censored
19
Perhapsbecausethisarticledripswithscornfortheconcept,andbaldydeclaresthat“asfarasoursocialistnationisconcerned,thecontroversybetweentheruleofmanandtheruleoflawhasnosignificancewhatsoever.”25“Asocialistlegalsystemisnotforshacklingthelimbsofthepeople.Itismainlydirectedatalloftheenemiesofthelaboringpeople.Wedonotregardlawastranscendingclass,orasdecidingeverythinginperpetuity,butrathergraspitasoneoftheprincipalmeansforeffectingthedictatorshipoftheproletariat.”26Clearly,thisisnotthesocialistruleoflawtraditiontoday’sauthoritiesarelookingfor.Whataboutthepeakarticleintheuncensoredclass?Thisarticleconsidersthedoctrineofthe“ruleoflaw”proposedbythethirdcenturyBCEphilosopherHanFei.27Hisisastern,top-downvisionoflawcloserintheWesterncanontoclassicalrechtsstaat,or“rulebylaw.”Itemphasizestheinterestsofthestateovertheindividual,andgrantsthesovereignunquestionedsupremacyoversubjects.Itssynergywithmodernauthoritarianismisevident,andasasourceoflegaltraditionwithindigenousrootsdeepenoughtoresisttheintrusionofmodernWesterncompetitors,theCCPcouldhardlydobetter.Hence,itnotonlyavoidsthescalpel,butdominatesthe(blue)sanitizeddiscourseonthe“ruleoflaw”remainingonline.
Figure14:RuleofLaw,Zhengfayanjiu(1956-58)
Bycontrast,inZhengfayanjiu,thepeakarticleremainsuncensored.(Figure14)Anaccountofafour-daystrugglemeetingorganizedinShanghaiatthestartoftheAnti-RightistCampaign,thearticleclaimedthat“atthemeetingmanypeopleforcefullyrefutedWangZaoshi’sslanderagainstthelegalsystem,tookonhisabsurdnotionof[China]‘stressingruleofmanwhile
25ZhaiTingjinÎÅÑ,“Boyoupaifenziderenzhizhuyimiulun”Ç�§�����)�Ëa[RefutingtheAbsurd
RightistElementTheoriesontheDoctrineoftheRuleofMan],FaxueM [LawScience]6(1957):8.26Zhai,“Boyoupaifenziderenzhizhuyimiulun,”11,43.27MouJieͲ,“LüelunHanFeidefazhizhuyi”�a�e�M�)�[IntroductiontoHanFei'sDoctrineofthe
RuleofLaw],FaxueM [LawScience]3(1957):9-15.
0
0.1
0.2
0.3
tf-id
fscore
Year.Issue
Uncensored
Censored
20
slightingtheruleoflaw’,andlistednumeroushardfactstoexplainthatChinahasthebestruleoflaw.LuoZhufengsaid:‘thequalityofruleoflawcanbemeasuredonthebasisofsocialorderanddemocraticrights,andinChinathestabilityofthesocialorderandtherealizationofdemocraticrightshashistoricallyneverbeengreater.’”28Thisisthenativesocialistruleoflawtraditionthecensorswishustodiscover,andtheyhavethoughtfullyclearedapathtoitbywinnowingawayallrivals.Thusfar,Ihavespokenonlyoftwolawjournalsfromanarrowsliceintime,andtheattentivereadermightwonderifthecaseIammakingappliesmorebroadly.Itdoes.Letuslookfirstatanotherlawjournalfromthesameperiod.Renminsifagongzuo�� "��[People’sJudicialWork]debutedin1957astheofficialorganofthePRCMinistryofJustice.Thefollowingyear,itmovedtotheSupremePeople’sCourtandwasrenamedRenminsifa�� "[People’sJudicature].Itscirculationwasrestricted(neibu)underbothsponsors,andthebattlesfoughtinitspages,whichmirroredthoseinZhengfayanjiuandFaxue,werethereforehiddenfrompublicview.Thiswaswherethejudiciaryspokeprivatelytoitself.Tothebestofmyknowledge,nolibraryintheUnitedStateshasphysicalcopiesofthe1957and1958volumesofthisjournal,whichmakesthePRConlineeditionprecious.Unfortunately,onlyoneofthefouronlinevendorsoffersfull-textarticlesfromanyofthesevolumes,anditscoverageisfragmentary.Allof1957ismissing,comprising12issues,or292articlesand513pagesoftext.Theonlineeditionalsoomits135articlesfrom1958,includingallofissues1-3and5-10,or40percentofthearticlesfortheyear.Thisvirtuallyzeroesouttherecordofthejournal’sparticipationintheHundredFlowersCampaignandtheoutbreakoftheAnti-RightistCampaign.Itisnothardtounderstandwhy.Manyofthemissingarticlespilloryhigh-rankingrightistsinthejudiciaryfortheir“anti-Party”attachmentstoprincipleslikejudicialindependence,theburdenofproof,andweighingtheargumentsfrombothsidesinacase.ThathistorynowembarrassestheCCP.Consider,forexample,thisexcerptfromapageoneeditorialinoneofthecensoredissues,againtheofficialorganoftheSupremePeople’sCourt:
Rightistswantedtouse“adjudicatoryindependence”toopposetheleadershipofthePartyoverjudicialwork…Isthisruleoflaworruleofman?Wehaveruleofman.WhatiscalledtheruleofmanisrelyingonPartyleadership,relyingonthemasses,givingfreereintothedynamismofhumansubjectivity.29
28“Shanghaifaxuejiezhankaifanyoupaifenzidedouzheng”�IM J}�n�§�����[Shanghai'sLegal
CommunityLaunchestheStruggleAgainstRightistElements],Zhengfayanjiu�M��[Political-LegalResearch]4(1957):53.29“Baxianchanghuiyidejingshendaidaoshijigongzuozhongqu”4�B ��Y\E3�O2��[Carry
theSpiritoftheThisMeetingtoPracticalWork],Renminsifa�boM[People'sJudicature]19(1958):4.
21
Fairenough,butonemightstillwonderifthecensorshipislimitedtolawjournals.Tobecertain,Icheckedissuesofaleadinginterdisciplinarysocialsciencejournal:Jiaoxueyuyanjiu���&'[TeachingandResearch].Thesamepairofonlineplatforms(CNKIandNSSD)offerfull-textcoverage,andtheirholdingsandomissionsareagainidentical.Nearly20percentofthetotalpagecountpublishedfrom1956through1958iscensoredonline,includingfivecompleteissuesandapproximatelytwodozenadditionalscatteredarticlesontopicsrangingfromphilosophytothearts,andvillagelife.(Figure15)Additionally,thepublisher,People’sUniversity,offersthejournalonitsownservers,andomitsthesameissues,whichwouldbeconsistentwithauniformcensorshipdirectivefromabove.30
Figure15:CensorshipAcrossOtherJournals,byPageCount
Jiaoxueyuyanjiu(1956-58) Faxueyanjiu(1978-79)
AskepticmightfurtherwonderifthecensorshipIhavefoundsimplyreflectstheParty’sspecialsensitivitiesabouttheAnti-Rightistperiod.Notatall.TheAnti-RightistCampaignismerelyoneexampleofamuchmoreextensiveproblem;theCCP’sinsecurityanddeepcontemptforhistoryguaranteethatnotopicorperiodissafefromsimilarmanipulation.Wehaveonlytoturntothereconstructionofthelegalsystemduringthelate1970sforproofofthat.ThefirstacademiclawjournaltoappearaftertheCulturalRevolutionwasFaxueyanjiu"�&'[StudiesinLaw],thesuccessortoZhengfayanjiu.Sevenissueswerepublishedfromlate1978totheendof1979,thefirsttwoofwhichweretrialballoonswithrestricted,internalcirculation.PublicationofthejournalcontinuestothisdayunderthesponsorshipoftheInstituteofLawoftheChineseAcademyofSocialSciencesinBeijing.
30Thejournalwebsiteis:http://xsqks.ruc.edu.cn/Jweb_jxyyj/CN/article/showOldVolumn.do
20.5%
79.5%
Censored Uncensored
87.2%
12.8%
Censored Uncensored
22
Atthetime,DengXiaopingwasstillconsolidatingpower.TocheckhisrivalsontheleftandtolaythegroundworkforthetrialoftheGangofFour,hepermittedlegaldiscourseadegreeoflatitudenotseensincethemid-1950s.ThiscoincidedwiththeDemocracyWallmovement,whichtookpopularpoliticalprotestanddissenttothestreets.The1978and1979issuesofFaxueyanjiudocumentthatbriefopening.Manyofthearticlesinthoseissuesreflectonthesorrystateofthelegalsystem,whichtheyblameontheGangofFour.Butsomehintatwiderculpabilityandexpressviewsthathavesincefallenoutoffavor.Forinstance,onearticlefranklyobserves:“itisobvioustoeveryonethatovertheyearsthedoctrineoftheruleofmanharmedChinesepoliticallife.Legalnihilismandthedoctrineoftheruleofmangainedgroundasearlyasthelate1950s.Throughtheearly1960s,thestandpointoffavoringtheruleofmanandrejectingtheruleoflawwasextensivelypropagated.”31AnotherarticleassertsthatLinBiaoandtheGangofFour“notonlyinheritedthedregsofhistoricalConfucianandLegalistthought,butalsopromotedthembyusurpingpower.Foratime,revolutionaryleadersweredeifiedandfeudal-fascismranwild,concealedundertheheavycurtainoftheruleofman.”32SuchopinionsmayhavebeenconvenientfortheproximateaimofdiscreditingtheGangofFour,buttheyareheterodoxtoday,andcensorshipprovidesatemptingwayoutofthatquandary.CensorshipfreesthePartytoargueinsteadthattherootsofsocialistruleoflawinthePRCrundeep,andthatConfucianandLegalistphilosophyareactuallyexcellentsourcesoflegaltraditionforit.33Iftheyhavedonetheirjobswell,nothingintheonlinerecordwilldisprovethis,orindicatethatorthodoxyoncedictatedotherwise.AcheckoftherecordrevealsthatonlyoneofthefourmajoronlinevendorsinthePRCoffersanyfull-textcoverageofFaxueyanjiu’s1978and1979issues,andtheholdingsarethin.Morethan87percentofthetotalpagecountismissing,includingthepiecescitedabove,andallsixissuesfrom1979.Forthesubscriber,itisalmostasifthediscoursethatjumpstartedacademiclegalscholarshipinthepost-MaoPRC,andcapturedthepromiseofthatpregnantmoment,neverexisted.(Figure15)Admittedly,norepositoryofthepastcanachievecompleteness,butsomethinginsidiouslurksbehindtheinsufficienciesoftheseparticularplatforms.Theyareexploitingthemalleabilityofthedigitalrecordtoturnhistoryonitshead,portrayingcurrentlegalreformasrestingona
31GuChundex�,LüShilunÈ=¦,andLiuXinª�,“Lunrenzhiyufazhi”a��7M�[OntheRuleofMan
andtheRuleofLaw],FaxueyanjiuM ��[StudiesinLaw]5(1979):27.32ZhangJinfanV¿Ì,andZengXianyitÉ�,“Renzhiyufazhidelishipouxi”��7M��~�Ä®[Historical
AnalysisoftheRuleofManandtheRuleofLaw],FaxueyanjiuM ��[StudiesinLaw]5(1979):37.33Zhonggongzhongyangzhengfaweiyuanhui,Shehuizhuyifazhilinianduben.
23
venerablefoundingvision.34Thisisoriginalismbycensorship,andwhileirksomeenoughforforeigners,itsprimaryaudienceissurelyChinese.ItaimstocontrolChina’sfuturebyshapingconsciousnessofitspast.Withthepassageoftime,detecting,exposinganddebunkingsuchploysgetseverharderandmoreexpensive,andthisalonemaybeenoughforthemtosucceed.Thehazardsdescribedhereundoubtedlyapplyeverywherethedigitalturntakeshold.Ifknowledgeispower,thentheconsolidationofinformationinthehandsofacomparativelysmallnumberofdigitalproviders,andtheconversionofourlibrariesfromredundant,fault-tolerantrepositoriesofphysicalobjectsintopassivelinksinacentralizeddistributionchaincontrolledbythoseprovidersisworrying.Increasingly,wearedependentonthegoodfaithoftheirstewardship,andvulnerabletothepolitical,regulatory,commercialandlicensingtermsthatmayimpingeuponit.Moreover,wearevoluntarilysurrenderingtheevidencenecessarytoindependentlymonitortheirperformanceandholdthemtoaccount.Icannotsayifthecensorshipthisstudyinvestigatesisalgorithmically-driven,buttoexplorethatpossibilityIusedamultinomialnaïveBayesclassifierthatemploysmachinelearningtobuildacensorshipmodel,whichIthentestedagainstaknowncontrolset.Thisresemblesthetechnologyusedtoidentifyandfilterspamemail.Themodelregularlyachievedmorethan70%accuracy,arespectablelevelgiventhesmallsizeofthecensoredclass,andthecommongenretowhichallthearticlesbelonged.Thesepreliminaryresults,performedonasmalldatasetwithamodestlaptopcomputer,showthatfromadistanceIcantrainamachinetoapproximatethechoicesthecensorsareactuallymakingaboutthesejournals,whichfurtherunderscoresthelikelihoodthatthereisadeterministic,discoverablelogicguidingtheirdecisions.Consideringthescaleoftheirresponsibilities,thecensorsmayalreadybeusingsimilartechnology.Withtheresourcesavailabletoastateortoawell-financedprivatefirm,thepossibilitiesarechilling.Forinstance,simplybyfine-tuninganyofthe30,000+parametersinmydataset,anenterprisingcensor(orhacker)couldfabricatebespokecorporawithsurgicalprecisionasoftenastheylike,withoutleavingbehindanyfingerprints,andsubscribinglibrarieswouldinturnfaithfullydisseminateeachsuccessiveedit.Likewise,Chinesestatearchives,manyofwhichareaggressivelydigitizingtheircollections,couldusesuchmethodstomodulatetheaccessiblerecord.Researcherswouldthenintegratethesedynamicallycuratedsourcebasesintotheirscholarshipandpropagatetheirbiasesaroundtheworld,muchasinaclassicdisinformationcampaign,onlyexecutedwithnegligibletransactioncostsandseamlessefficiency.
34Formoreonthistopic,seeGlennTiffert,“SocialistRuleofLawWithChineseCharacteristics:ANewGenealogy,”inSocialistLawinSocialistEastAsia,ed.HualingFu,JohnGillespie,PipNicholson,andWilliamPartlett(NewYork:CambridgeUniversityPress,forthcoming).
24
Solutions:Oneobviousresponsewouldbetobringthemissingtextstotheattentionofthevendors,andrequestremediation.Myeffortsasanindividualconsumermetwithdisappointment.ButasubscriptiontooneoftheseChinesedigitalplatformscancostaresearchuniversitytensofthousandsofdollars,andthepossibilityofcollectiveactionbyagroupofthemtodemandabetterproductcouldprodthevendorsintherightdirection.Atthesametime,theevidencealsosuggeststhatthedecisionmaynotbetheirstomake,andpersuadingtheChinesegovernmenttoreverseitselfisamuchtallerorder.Anotherresponsewouldbetorepublishthecensoredtexts,butthereisavexingobstacletothat.UndercurrentU.S.law,allofthejournalsexaminedinthisstudyareundercopyrightandwillremainsountil95yearsaftertheirdateoforiginalpublication,i.e.atleastuntiltheearly2050s.35Thisforeclosesmanyoftheoptionsonecantaketocircumventtheircensorshipsinceanyunauthorizedrepublicationofthesourceswouldconstituteinfringement.Aswithconventionallyorphanedworks,fewinstitutionswouldriskliability,andtheChineserightsholderscoulduseU.S.lawasaclubtoenjoinanythatdid.Thus,intellectualpropertylawperverselyabetsthemanipulationofhistorybyanauthoritarianregime.Thatregimeneedonlywaitforthefewsurvivingphysicalcopiesofthecensoredtextstocrumbleorfadeintoobscurity.Manylibrariesareactuallyacceleratingthoseoutcomesbydeprecatinglegacyvolumesontheirowninitiative,oblivioustothestakes.Infact,allcopyrightedworksaresusceptibletothismorbidwaitinggame.Forexample,theiconicfilm,RiverElegy( �HeShang),bannedinthePRCsince1989,languishesinmuchthesameposition.AconcertedeffortoutsideofChinatodigitizethesematerialsforthesakeofpreservation,andtosequesterthemuntiltheyfalloutofcopyrightdecadesfromnowisperhapsthemoststraightforwardsolution,thoughadeeplyunsatisfyingone.Technologyhasotherpartstoplay.Inthisstudy,Iusedcomputationaltoolstodemonstratethatthedystopianpossibilitiesofthedigitalturnarealreadyamongus.Thereisacertainironyinthat.Thesetoolsequippedmetoanalyzetherelationshipsamongdocumentsatascaleandresolutionbeyondmyowncognitiveabilities.Butratherthanusethemasreplacementsfortraditionalmethods,Iemployedthemcomplementarily--asinstrumentsforhypothesistesting,topullinterestingfeaturesoutofmycorporaforcloserinterpretivescrutiny,andtocheckmysubjectivejudgmentsempirically.Thetoolswerealsoinvaluableforanalyzingthealternatehistoriescreatedbythecensors,assessingwhathasbeenlost,distorted,andfabricatedbytheirwork,reverseengineeringitslogic,andreconstitutingtherecordtheymanipulated.Inshort,technologyelevatedmyconventionalskillsasahistorian.3517U.S.C.§104A;ElizabethTownsendGard,“VeraBrittain,Section104(a)andSection104A:ACaseStudyinSortingOutDurationofForeignWorksUnderthe1976CopyrightAct,”7thAnnualIntellectualPropertyScholarsConference,DePaulUniversityCollegeofLaw(2007):n.p.;TylerT.Ochoa,“CopyrightProtectionforWorksofForeignOrigin,”inTheInternationalizationofLawandLegalEducation,ed.JanKlabbers,andMortimerSellers(London:Springer,2008):167-190.
25
Itcanalsopreventthecancerthisstudyhasexposedfromspreading.Already,thedigitalturnissparkingcynicisminsocietyabouttheauthenticityoftheknowledgeweconsumeandproduce.Amerewhiffclosertohomeofthepracticesthisstudyhasdocumentedcouldenflamethatcrisisofcredibilityandfurthererodethetrustnecessaryforrobustscholarshipanddemocraticpractice.36Itisthereforeincumbentonknowledgecreators,rightsholders,digitalprovidersandtheinstitutionsthatsubscribetotheirproductstodesignandconsistentlyimplementasetofbestpracticestoupholdtheintegrityoftheirholdings,transparentlylogomissionsandmodifications,anddefendagainsttamperingatthelevelsoftheindividualcharacter,documentandcorpus.Wecannolongerpassivelytrust,wemustalsoverify.Avarietyofsolutions,suchasdigitalsignaturesorblockchaincertification,withlogossignifyingvalidatedstandardscompliance,arepotentiallyavailabletomeetthoseneeds.37Thedangerisreal.Neverbeforehasknowledgebeenpronetosuchsweepingandsupplemanipulation.Historyhangsinthebalance.
36SamuelC.Woolley,andPhilipN.Howard,“ComputationalPropagandaWorldwide:ExecutiveSummary.”ComputationalPropagandaResearchProjectWorkingPaperNo.2017.11(2017):accessedJune21,2017,http://comprop.oii.ox.ac.uk/2017/06/19/computational-propaganda-worldwide-executive-summary/.EnvironmentalDataandGovernanceInitiative.“TheEPAUnderSiege.”(2017):AccessedJune30,2017.https://100days.envirodatagov.org/epa-under-siege.html.37Portico(http://www.portico.org),anon-profit,digitalpreservationservicethatpartnerswithlibrariesandpublisherstopreservethescholarlyrecord,offersapromisingpartialsolutionforcommitteddigitalcontentnolongeravailablefromthepublisheroranyothersource,butitsstandardlicensingagreementsdonotsafeguardagainstthefullrangeofpracticesdescribedinthisstudy,particularlycensorshipatthelevelofindividualarticles.
26
Appendix:ToolsandMethodsPreparingthedocumentsfortheanalysisundertakeninthisstudyrequiredseveral,labor-intensivesteps,themostbasicofwhichwasconvertingimagesintotextwithahighdegreeofaccuracy.38Tomaximizefidelity,Iusedonlyoriginalprinteditions,notreprintsorreproductions.Everypageofeveryissueinthethree-yearsample,morethan2,000intotal,wasscannedat600dpi,ingreyscalewithoutanylossycompression,resultinginnearly20gigabytesofdata.Second,Iusedacommercialopticalcharacterrecognition(OCR)packagetoconvertthosescansintoplaintext,andIthenslicedtheoutputintoindividualfiles,oneforeacharticle.Myfinalcorporaconsistedof356filesfromZhengfayanjiuand381filesfromFaxue,ornearlyfourmillioncharactersintotal.39Bothvendorsarecurrentlycensoringapproximatelyelevenpercentofthetotalpagecountinmythree-yearsample,thoughthatshareexceeds50percentforcertainissuesstraddlingthe1957/58divide.Third,university-educated,nativespeakersofChinesecomparedmyFaxuecorpusagainsttheoriginaldocuments,characterbycharacter,toestablishthereliabilityoftheOCR.Thetestsetcomprised127pages(approximately207,000characters)fromtheoriginalissues,andaveragedaremarkable99percentaccuracy,easilysufficientformypurposes.TimeandfundingprecludedanequallyexhaustivetestofmyZhengfayanjiucorpus,butspotcheckssuggestedsimilaraccuracy.Fourth,IwroteseveralprogramsinPython,thefirstofwhichstrippedmycorporaofsemanticallyexpendablecharacters,suchaspunctuation,alphanumerics,Cyrillic,gremlins,andwhitespace.ThisreducedeachfiletoanunbrokenstreamofChinesecharacters,whichwasthenfedtoasegmentationalgorithmthattokenizedthetexts.UnlikeWesternlanguages,Chinesedoesnotseparatewordswithwhitespace.Thesegmentationalgorithmperformsthisstep,whichisrequiredbythenaturallanguageprocessing(NLP)routinesthatcompriseconventionalcomputationaltextanalysis.40ItbearsmentioningthatthetextsinmycorporawereoriginallypublishedjustasthePRCwasmovingfromtraditionalChinesecharacterstothe
38Myworkflowreliedonthefollowingsoftware:MacOS(10.12.5),ABBYYFineReader(12.1.6),BBEdit(11.6.5),Anaconda3(3.4.1),Scikit-learn(0.18.1),Java8(131),StanfordChineseSegmenter(3.7)withthectbstandard,Mallet(2.08),andMicrosoftExcel(15.35).39Onecouldarriveataslightlydifferentcount,dependingonhowonedividessidebarsandforumswithcontributionsfrommultipleauthors,butthekeypointisthateverycharacterwascaptured.40Mymethodologyandworkflowwereinformedbyvarioustextsincomputationallinguistics,naturallanguageprocessing,anddigitalhistory,including:CharuC.Aggarwal,andChengxiangZhai,eds.MiningTextData(NewYork:Springer,2012);StevenBird,EwanKlein,andEdwardLoper,NaturalLanguageProcessingWithPython(Cambridge:O’Reilly,2009);MichaelC.Hout,MeganH.Papesh,andStephenD.Goldinger,“MultidimensionalScaling,”WileyInterdisciplinaryReviewofCognitiveScience4,no.1(2013):93-103;DavidMimno,“ComputationalHistoriography:DataMininginaCenturyofClassicsJournals,”ACMJournalofComputinginCulturalHeritage5,no.1(2012):3:1-3:19;JasonD.M.Rennie,etal.,“TacklingthePoorAssumptionsofNaiveBayesTextClassifiers,”ProceedingsoftheTwentiethInternationalConferenceonMachineLeaning(2003):616-623;DavidUnderhill,etal.,“EnhancingTextAnalysisViaDimensionalityReduction,”IEEEInternationalConferenceonInformationReuseandIntegration(2007):348-353.PortionsofmycodeadaptedexamplessharedbyPaulVierthaler,BrandonRose,andcountlesscontributorstoStackOverflow.
27
simplifiedsetusedtoday,andthearticlesconsequentlycontainatransitionalmixtureofboth.Thisconfusedthesegmentationalgorithm,andnecessitatedthepreliminarymeasureofconvertingbothcorporauniformlytosimplifiedcharacters.Fifth,thesegmentationalgorithmperformedpoorlyfornamedentities,particularlypersonalnames,whichcanbeidiosyncratic.Forexample,surnameswerecommonlydissociatedfromgivennames.Ithereforemanuallycompiledadictionaryofseveralhundredentriescomprisingeveryauthor,aswellasprominentorganizationsandindividualsmentionedinmytexts.APythonprogramthenscannedthetextsforoccurrencesofthesenamesandreconstitutedthemcorrectly.Sixth,IcompiledaseriesofmetadatafilesnecessaryfortheanalyticsIintendedtoperform.Thesefilesareessentiallyspreadsheets(.csv)withcolumnsforthefilenames,articletitles,authornames,indexnumber(year-issue#-article#),andcensorshipstatus(YorN)ofeveryarticleinvariousslicesofmycorpora.IbuiltmetadatafilesfortheZhengfayanjiucorpus,theFaxuecorpus,acombinedcorpus,andyetanothercombinedcorpusthatsubstitutedthecityofpublication(B[eijing]orS[hanghai])forthecensorshipfield.Finally,afterapproximatelyfivemonthsofpreparation,thecorporawerereadyforanalysis.Toperformtheanalytics,IwroteanotherPythonprogramthatstrippedoutstopwordsfromeachpre-processedcorpusandthentransformedthemintoaseriesoftf-idfmatricesagainstwhichIcouldrunabatteryofstatisticaltests.41Briefly,thesematricesdescribeeachofthe737documentsinmycorporainmorethan30,000dimensions,whereeachdimensioncorrespondstoauniqueterminadocument(e.g.“judge,”“YangZhaolong,”“ruleoflaw,”etc.).Forthegraphs,Ithenappliedavarietyofstandardtechniquestoreducethatspacetotwoorthreeabstracteddimensionsthatnonethelessretainsufficientinformationaboutthedocumentstorepresenttheiressentialrelationshipsinamannerthathumanscanvisuallycomprehend.42
41Bydefinition,stopwordsdonotcarryinformationsignificanttoagivenquery,usuallybecauseoftheirhighfrequency.ThearchetypicalexampleinEnglishwouldbe“the.”Filteringthemoutremoveschafffromthedataandreducesprocessingtime,withoutmateriallyaffectingresults.Atermfrequency-inversedocumentfrequency(tf-idf)matrixisastandardstatisticalrepresentationofacorpus,inwhicheachtermineverydocumentisassignedaweightedvaluebasedonitsfrequencyinthatdocumentanditsdistributioninthecorpusatlarge.Thebasicideaisthatatermthatappearsofteninasmallfractionofthecorpusismoreinformativeandofgreaterselectionvaluethanatermthatisdistributedevenly.Thenormalizedtf-idfvalueswerecomputedusingtheTfidfVectorizerclassinScikit-learn0.18.1,accordingtothefollowingfunction:!"#$"(&,() = !"(&,()×,-. /012
/0(3(2,4)+ 1.
42Thesetechniquesincludedt-SNEanalysis(TruncatedSVD),principalcomponentanalysis,Euclideandistancecalculations(MDS),screeplots,cosinesimilaritycalculations(MDS),sphericalk-meansclustering(MDS),hierarchicalclusteringanalysis(Warddendrogram),andnaïveBayesclassification(multinomial).Ialsoperformedχ2featureselection,andtopicmodeling(Mallet,LDA)onmycorpora.
Related Documents
![L’histoire par la photographiecentregeorgsimmel.ehess.fr/files/2016/01/About-Chéroux...n’ait pas été inventée plus tôt. Arlette Farge déplore ainsi que « personne [n’ait]](https://static.cupdf.com/doc/110x72/609c6da1bff01841e00cb866/lahistoire-par-la-photograp-roux-naait-pas-t-invente-plus-tt-arlette.jpg)
