1-5 17.选对老师，智慧一生；选对伴侣，幸福一生；选对朋友，受益一生；选对行业，成就一生。面对朋友，不可欺骗；面对家庭，不可背叛；面对老婆，不可逾越。人，上半生要不犹豫，下半生要不后悔。活在当下，把握每次机会，因机会稍纵即逝。
1-5 16.【德鲁克经典语录】1、企业走下坡路的第一个信号，是对那些合格的，能干的和有志向的人才，失去吸引力；2、管理不在于 “知”，而在于“行”；3、我永远不会提拔一个从不犯错误、特别是从不犯大错误的人担任最高层的工作.
1-5 15.【人生语录】1不懂装懂是聪明，懂装不懂是智慧。2示弱而不逞强，示拙而不逞能。3忍人之所不能忍，方能为人之所不能为。4身做好事，言说好话，心存好念。5大悲无泪，大悟无言，大喜无声，大爱无言。6君子相交，随方就圆，无处不自在。7智者不锐，慧者不傲，谋者不露，强者不暴。
1-5 15.人生三修炼：看得透想得开，拿得起放得下，立得正行得直。人生三福：平安是福，健康是福，吃亏是福。人生三不争：不与上级争锋，不与同级争宠，不与下级争功。人生三大快事：美酒，挚友，枕边书。 人生三大憾事：遇良友不交，遇良机不握，遇老师不学。 人生三为：和为贵，善为本，诚为先。
1-5 14.【比尔盖茨的忠告】1社会充满不公平，你先不要想去改造它，只能先适应它。2世界不会在意你的自尊，人们看到的只是你的成就。在没有成就之前，切勿过分强调自尊。3当你陷入困境时，不要抱怨。默默地吸取教训。4学校里有节假日，到公司打工却不尽然。你几乎不能休息。很少能轻松过节假日。
1-5 14.改变命运的经典途径：1、比大多数同事勤奋认真。2、做好每一件小事。3、以微笑面对任何命运的不公。4、时间比金钱重要。5、养成好习惯，习惯决定命运。6、不断与人性弱点斗争。7、天才都是有心人。8、机会无大小，只有时间早晚；永远不嫌弃小机会。
9、等待大机会是痴心妄想。10、坚持不懈锻炼身体。
1-5 13.一年轻人觉得自己怀才不遇，有位老人听了，随即把一粒沙子扔在沙滩上，说：“请把它找回来。”“这怎么可能！”接着老人又把一颗珍珠扔到沙滩上，“那现在呢？”如果你只是沙滩中的一粒沙，那你不能苛求别人注意你，认可你。如果要别人认可你，那你就想办法先让自己变成一颗珍珠。
1-5 12.【忍与让】一味忍让，意味着丧失原则;一味忍让，意味着没有人格;一味忍让， 意味着软弱可欺;一味忍让，意味着面临步步进逼的危险;一味忍让，意味着将走入绝路。有时候挺身而起、奋力反抗效果更好，得寸进尺是愚人常采用的计策，一再忍让反而助长其嚣张气焰。该出手时就出手，给点厉害也是不得已而为之!
1-4 11.十大奢侈品: 1、一颗不老的童心。2、生生不息的信念。3、背包走天下的健康。4、愉悦心情与性情的工作。5、安稳与平和心绪的睡眠。6、享有属于自己空间与时间的生活。
7、牵手一个教会你爱与被爱的人。8、品味美丽和美好的心与心情。9、自由的心态与宽广的胸襟。10、点燃他人希望的精神特质。
1-4 10.从自己做起：1、只有你爬到山顶了，这座山才会支撑着你；2、只有你的境界提高了，这个境界才来提升你；3、只有你关心别人了，别人才会关心你；4、只有你爱护这个环境了，这个环境才会爱护你，6、只有你的亲人幸福了，你才会幸福；6、只有你自己成功了,朋友才会离你更近。
1-3 9.【八种话不能讲】一、丧志的话不能讲；二、负气的话不能讲；三、抱怨的话不能讲；四、损人的话不能讲；五、自夸的话不能讲；六、不实的话不能讲；七、机密的话不能讲；八、隐私的话不能讲。讲话就像泼水，泼出去的水无法再收回。
1-3 8.【送给所有美人们】1.喜欢的东西自己努力买，不要指望别人送。2.可以淘便宜的衣服，但记得自己的品位比这个价位高。3.在QQ，手机里删除前男友的号码，避免神经脆弱
的时候主动找他。4.减肥是为了更美好的人生。要是因为减肥而失去了生活的乐趣，不如放弃。5.注重内心，但不忽略外表。
1-3 7.【优秀女人的潜质】1、执著但不固执。2、随和但不随性。3、出色但不出格。4、低调但不低俗。5、痴情但不矫情。6、自信但不自我。7、诱人但不缠人。8、计算但不算计。
9、幽默但不油滑。10、潮流但不风流。11、忍耐但不隐忍。12、飘逸但不飘荡。
1-3 6.【心理学之做人六字诀】1.静：少说话，多倾听。2.缓：稳着做事，不急不躁。3.忍：面对不公，别气愤，别宣泄，忍让是智慧。4.让：退一步，海阔天空。5.淡：一切都看淡些，很多事情随着时间会变成云烟。6.平：是平凡，是平淡，是平衡
1-2 5.【提升自我幸福感的五种做法】1、经常联系周围的人。你的家庭、朋友、同事还有邻居。2、让身体动起来去散散步或者跑一圈。3、留心周围，留心那些不经意的美丽。4、学无止境，尝试一些新鲜的东西或找回你过往的某一种兴趣爱好。5、与人为善，尽量多给朋友甚至陌生人提供帮助。
1-1 4.【禅语悟道】1人生有八苦：生，老，病，死，爱别离，怨长久，求不得，放不下。2命由己造，相由心生，世间万物皆是化相，心不动，万物皆不动，心不变，万物皆不变。3笑着面对，不去埋怨。悠然，随心，随性，随缘。注定让一生改变的，只在百年后，那一朵花开的时间。
1-1 3. 1、 人与人的最大差别，不在于脖子以下90%的体力劳动，而在于脖子以上仅占10%的智力。2、人生如同故事，重要的并不在有多长，而是在有多好。3、人真正的完美不在于他拥有什么，而在于他是什么。4、做噩梦的好处是，你醒来后发现生活还不算太坏。 1-1 2.【快乐心情六方法】①走一走，散散步，让自然荡涤心胸；②比一比，阿Q精神胜利法，比上不足比下有余；③放一放，不是急事大事索性放下，回头再说；④乐一乐，想想开心事，看一看笑话 ；⑤唱一唱，回忆过去似水年华，温暖你自己；⑥让一让，人生如狭路行车，眼光放远，让一步。
1-1 1.【建立人脉的10大潜规则】1、想钓到鱼,就要像鱼那样思考；2、不要总显示比别人聪明；3、让对方作主角,自己甘愿做配角；4、目中无人,让你一败涂地；5、常与人争辩,你永难赢；6、锋芒太露,下场不好；7、刺猬原则-保持适当距离；8、树一个敌,等于立一堵墙；
9、谦虚不虚伪,不苛求完美；10、失言不如无言。
第二篇：德鲁克积分
Edited by Foxit ReaderCopyright(C) by Foxit Software Company,2005-2007For Evaluation Only.
Folksonomies-CooperativeClassi?cationandCommunicationThroughSharedMetadata
AdamMathes
ComputerMediatedCommunication-LIS590CMC
GraduateSchoolofLibraryandInformationScience
UniversityofIllinoisUrbana-Champaign
December2004
Abstract
Thispaperexaminesuser-generatedmetadataasimplementedand
appliedintwowebservicesdesignedtoshareandorganizedigitalme-diatobetterunderstandgrassrootsclassi?cation.Metadata-dataaboutdata-allowssystemstocollocaterelatedinformation,andhelpsusers?ndrelevantinformation.Thecreationofmetadatahasgenerallybeenapproachedintwoways:professionalcreationandauthorcreation.Inli-brariesandotherorganizations,creatingmetadata,primarilyintheformofcatalogrecords,hastraditionallybeenthedomainofdedicatedprofes-sionalsworkingwithcomplex,detailedrulesetsandvocabularies.Theprimaryproblemwiththisapproachisscalabilityanditsimpracticalityforthevastamountsofcontentbeingproducedandused,especiallyontheWorldWideWeb.Theapparatusandtoolsbuiltaroundprofessionalcatalogingsystemsaregenerallytoocomplicatedforanyonewithoutspe-cializedtrainingandknowledge.Asecondapproachisformetadatatobecreatedbyauthors.Themovementtowardscreatordescribeddocu-mentswasheraldedbySGML,theWWW,andtheDublinCoreMetadataInitiative.Thereareproblemswiththisapproachaswell-oftenduetoinadequateorinaccuratedescription,oroutrightdeception.Thispaperexaminesathirdapproach:user-createdmetadata,whereusersofthedocumentsandmediacreatemetadatafortheirownindividualusethatisalsosharedthroughoutacommunity.
1TheCreationofMetadata:Professionals,Con-
tentCreators,Users
Metadataisoftencharacterizedas“dataaboutdata.”Metadataisinformation,oftenhighlystructured,aboutdocuments,books,articles,photographs,orotheritemsthatisdesignedtosupportspeci?cfunctions.Thesefunctionsareusually
1
Edited by Foxit ReaderCopyright(C) by Foxit Software Company,2005-2007For Evaluation Only.
tofacilitatesomeorganizationandaccessofinformation.Administrative,struc-tural,anddescriptivemetadataarethreebroadcategoriesofmetadata(Taylor,2004).Thispaperfocusprimarilyondescriptivemetadatawhichidenti?esandfunctionstoorganizeinformationbasedonitsintellectualcontent.
Traditionallymetadataiscreatedbydedicatedprofessionals.Catalogerscreatemetadata,oftenintheformofMachine-ReadableCataloging(MARC)recordsforbooksandotherintellectualcreations,andthisisthebasisofmostOnlinePublicAccessCatalogs(OPAC)inlibrariesandotherinstitutions.Thisoftenrequiresseriouseducationandtraining.Thelibraryandinformationsci-ence?eldhasdevelopedelaboraterulesandschemesforcataloging,categoriza-tionandclassi?cationthatincludeclassi?cationschemessuchastheDeweyDecimalSystemandLibraryofCongressClassi?cationScheme,aswellaslargecontrolledvocabulariesoftermsfordescribingthesubjectofmaterials,suchastheLibraryofCongressSubjectHeadings.
Whileprofessionallycreatedmetadataareoftenconsideredofhighquality,itiscostlyintermsoftimeande?orttoproduce.Thismakesitverydi?culttoscaleandkeepupwiththevastamountsofnewcontentbeingproduced,especiallyinnewmediumsliketheWorldWideWeb.Analternativeisauthorcreatedmetadata.Originalcreatorsoftheintellectualmaterialprovidemeta-dataalongwiththeircreations.TheDublinCoreMetadataInitiativehasbeenusedwithsomesuccessinthisarea(Greenbergetal,2002).Authorcreatedmetadatamayhelpwiththescalabilityproblemsincomparisontoprofessionalmetadata,butbothapproachesshareabasicproblem:theintendedandunin-tendedeventualusersoftheinformationaredisconnectedfromtheprocess.Usercreatedmetadataisathirdapproach,andthispaperfocusesongrass-rootscommunityclassi?cationofdigitalassets.Otherformsofusercreatedmetadataareoftenimplicit.Citationanalysisisawellestablishedtechniqueusedtodeterminerelationshipsbetweenacademicworksandtheimpactofschol-ars.SimilaranalysisofthelinkstructureintheWorldWideWebisusedbythePageRankalgorithm,whichbecamethetheoreticalbasisfortheGooglesearchengine(Page,1998).Recommendationsystems,andthosethatemploycollab-orative?lteringareanotherformofleveragingimplicitusercreatedmetadata.(Lieberman,2002).
Oneformofexplicitusercreatedmetadatawaspopularizedinthelate1990’swithlink-focusedwebsitescalledweblogs(Blood2000).Thesesitesprovidelinkscombinedwithcommentary,andareaformofmostlyunstructured,butexplicit,usercreatedmetadata.CustomerreviewsonwebsitessuchasAmazon.comareanintegralaspectofonlinecommerce,andleveragesconsumercreatedmetadatatocreatesitesthatarefarmoreinformativethancomparablecommercialsites.2TaggingContentinDel.icio.usandFlickr
Del.icio.us(http://del.icio.us,henceforthreferredtoas“Delicious”)isatooltoorganizewebpages.Adescriptiononlinestatesitis:
“asocialbookmarksmanager.Itallowsyoutoeasilyaddsitesyouliketo
2
Edited by Foxit ReaderCopyright(C) by Foxit Software Company,2005-2007For Evaluation Only.
yourpersonalcollectionoflinks,tocategorizethosesiteswithkeywords,andtoshareyourcollectionnotonlybetweenyourownbrowsersandmachines,butalsowithothers”(Schachter,2004)Deliciousisnotuniqueorpioneeringinitsroleasbookmarksmanager.Whatseemstoberelativelynewanddi?erentistheemphasisonuseraddedkeywordsasafundamentalorganizationalconstruct.Thesekeywords,whicharereferredtoas“tags”onthesite,allowuserstodescribeandorganizecontentwithanyvocabularytheychoose.
Tousethesystem,youmust?rstjoinbyregisteringanaccount.Thesystemisfreetojoinanduse.Onlyausername,fullname,andpasswordarerequired.Theuserthenaddsaspecializedbookmarktotheirwebbrowser.Whenbrows-ingawebpagetheywouldliketoaddtodelicious,theyselectthebookmark,andarepresentedwithaformthathasallowsthemtoenteranytagstheywanttoassociatewiththepage,andthensaveit.Thesetagsareoptional;userscananddousethesitewithouttaggingtheirdocuments.
Inadditiontoautomaticallygeneratedchronologicalorderingofbookmarkssavedtothesystem,thetagsareusedtocollocatebookmarkswithinauser’scol-lection.Additionally,thesetagsarealsousedtocollocatebookmarksacrosstheentiresystem,soforexample,lookingatthepagehttp://del.icio.us/tag/linuxwillshowallbookmarksthataretaggedwith“linux”byanyuser.
Flickr(http://www.?ickr.com),aphotomanagementandsharingwebap-plication,hasasimilarsystemoffree-formtaggingforphotosthatwasadoptedandmodeledafterDelicious.Ittoorequiresuserstocreateauseraccount,andisfreetojoin.Thereisalsotheoptiontopayforanaccountwithmorefeatures,likemorestoragespaceforphotographs.Flickro?ersasimilarbookmarktoaddphotographstothesystem,butalsohasanumberofotheroptionstouploadphotographstothesystemthroughwebpagesandsoftwareapplications.Tagscanbeaddedatthetimeofupload,orlaterintheprocesswhenthephotographsaredisplayedbythesystem.
Aprimarydi?erencebetweenDeliciousandFlickristhatwhilethetagsonDeliciousareprimarilyfromtheusersofwebdocumentsthatwerewrittenbyanotherparty,Flickrisprimarilyusedbyindividualstomanagetheirowndigitalimages,andthemajorityofthetagsareuserstaggingphotostheycreatedthemselves.Thisisnotabsolute;thesystemdoeshavetheoptionofallowingusersdesignatedasfriendsorfamilytotagausers’photos.Additionally,userscananddoenterimagesotherscreatedintothesystem,oftenfromwebsites.ThisuseofthesystemismuchmorelikeDelicious,butseemstobeasmallfractionoftheuse.
3FromTagstoFolksonomy
ThemostpopulartagsusedonDeliciousarelistedontherightsideofthefrontpage.Relatedtags,asdeterminedprogrammaticallybythesystem,arelistedontherightsideofindividualtagpages.TheorganicsystemoforganizationdevelopinginDeliciousandFlickrwascalleda“folksonomy”byThomasVander
3
Walinadiscussiononaninformationarchitecturemailinglist(Smith,2004).Itisacombinationof“folk”and“taxonomy.”
Animportantaspectofafolksonomyisthatiscomprisedoftermsina?atnamespace:thatis,thereisnohierarchy,andnodirectlyspeci?edparent-childorsiblingrelationshipsbetweentheseterms.Thereare,however,automaticallygenerated“related”tags,whichclustertagsbasedoncommonURLs.Thisisunlikeformaltaxonomiesandclassi?cationschemeswheretherearemultiplekindofexplicitrelationshipsbetweenterms.Theserelationshipsincludethingslikebroader,narrower,aswellasrelatedterms.Thesefolksonomiesaresimplythesetoftermsthatagroupofuserstaggedcontentwith,theyarenotapredeterminedsetofclassi?cationtermsorlabels.
InDelicious,acursoryanalysisofthetagsrevealsthatthemostpopulartagsareprimarilysubjectdescriptorkeywordsatvariouslevelsofspeci?city.Someofthemostpopulartags(asofNovember14,2004)accordingtothesystemwere:“software,design,programming,music,politics,web,news,blog,css,linux,art,osx,java,mac,blogs,reference,fun,python,games,tech,photography,humor,tools,delicious,rss,?refox,toread,comics.”Manyofthesearetechnicalsubjecttagsre?ectingthecommoninterestsofatech-savvyuserbase,e.g.“rss,?refox,python,java,linux.”Somearebestdescribedasgenreorformdescriptors,like“comics,humor,fun,photography.”Atleastone,“toread,”issomethingquali-tativelydi?erent:itisatagapparentlyusedforself-organizationandreminder.Similarly,“wishlist”(http://del.icio.us/tag/wishlist)wasapparentlyusedbyanumberofuserstohighlightconsumeritemstheywereinterestedin.
The150mostpopulartagsonFlickraretabulatedandlistedonthesite.AsofNovember19,2004,thislistincludedmuchofwhatonemightexpectascommonsubjectsofphotos:cat,friends,dogsky,sea,park,kids,garden,baby,building,?ower,?owerssigns,sculpture,city,vacation.Over25%(41outof150)ofthetagswereproperplacenameslikecitiesorcountries.Colorswerelisted:yellow,green,blue,pink,orange,white,red.Yearswerealsopopularastags,2001,2002,2003and2004werepresentamongstthemostpopular.
Sometermsthathaveparticularmeaninginthephotographicdomainlikeportrait,macro,landscape,blackandwhitewereincluded.Theterms“camer-aphone,moblog,fotolog”re?ecttheuseofrelativelynewwords,andthecon-nectionthesitehastotechsavvyearlyadoptersofintegratedcameraphonetechnologyandweblogsfocusedonphotographs.
Twotagsofparticularinterestare“cute,”and“me.”Thiswillbeelaborateduponlater,butIthinkthesetwotermsre?ectthedualnatureofthesesystems:thecompulsiontoshare-whatistheInternetifnotavenueforsharingcutephotographs?-andconverselytheimportanceofindividualityandegoforthesesystemstowork.
Overall,althoughtheterm“classi?cation”isoftenusedinrelationtothesesystems,andhasbeenusedinthispaper,whatisgoingonismorelike“catego-rization.”Categorizationisgenerallylessrigorousandboundariesarelessclear.Itisbasedmoreonasynthesisofsimilaritythanasystematicarrangementofmaterials(Jacob2004).Mostimportantly,eachdocumentcanhavemanytermsassociatedwithit.Bycontrast,classi?cationschemesgenerallyfocusonpro-
4
vidingasingleclassi?cationtoanitem,andareveryhierarchicalandhaveclearrelations.Inafolksonomythesetoftermsisa?atnamespace:therearenoclearlyde?nedrelationsbetweenthetermsinthevocabulary.
3.1Limitations
Theproblemsinherentinanuncontrolledvocabularyleadtoanumberoflim-itationsandweaknessesinfolksonomies.Ambiguityofthetagscanemergeasusersapplythesametagindi?erentways.Attheoppositeendofthespectrum,thelackofsynonymcontrolcanleadtodi?erenttagsbeingusedforthesameconcept,precludingcollocation.
3.2Ambiguity
Asanuncontrolledvocabularythatissharedacrossanentiresystem,thetermsinafolksonomyhaveinherentambiguityasdi?erentusersapplytermstodoc-umentsindi?erentways.Therearenoexplicitsystematicguidelinesandnoscopenotes.Forexample,itemstaggedwith“?ltering”onDeliciousincludedthefollowing:
?Last.FM-Yourpersonalmusicnetwork-Personalizedonlineradiostation?InfoWorld:Collaborativeknowledgegardening
?Wired12.10:TheLongTail
?OhMyGodItBurns!“PracticalApplicationsofthePhilosopher’sstone.Fordrunks.Brita?ltermakesbadvodkaintogoodvodka
?IntroductiontoBayesianFiltering
Theseareall“?ltering,”butinverydi?erentsenses.Usingwater?lterstopurifyvodkaisaverydi?erentsubjectthanBayesianstatisticalanalysis.
Acronymspresentanotherareaofpotentialambiguitythatareoftendealtwithe?ectivelyincontrolledvocabularies.ExaminingthefrontpageonNovem-ber14,2004revealedoneusertaggingsiteswith“ANT.”AfterexaminingtheothersitestheusertaggedwithANT,itwasapparentthiswasanacronymfor“ActorNetworkTheory,”inthedomainofsociology.However,whenexaminingtheANTtagacrossallusers(Deliciousapparentlyisnotcasesensitiveintags)mostofthebookmarkswereaboutApacheAnt,aprojectbuildingtoolintheJavaprogramminglanguage.Twocompletelyseparatedomainsandideasaremixedtogetherinthesametag.
3.3Spaces,MultipleWords
BothDeliciousandFlickrseemdesignedprimarilytodealwithsinglewords.Deliciousdoesnotallowspacesintagnames,althoughFlickrdoes.Insomeinstances,multiplewordsareusedtogetherinasingletag,withoutspaces,
5
i.e.,vertigovideostillsbbc’onFlickr.Attimesthiscanre?ectuserstryingtoputahierarchyintoasingletag,orsimplyre?ectsacategorythathasmulti-pleterms,suchasdesign/css’onDelicious.(http://del.icio.us/tag/design/css.)Bothsystemsignorelettercase,whichmaycollapsedistinctideasintoasingletag,especiallywithacronyms.
3.4Synonyms
Thereisnosynonymcontrolinthesystem.Thisleadstotagsthatseeminglyhavesimilarintendedmeanings,like“mac,”“macintosh,”and“apple”allbeingusedtodescribematerialsrelatedtoAppleMacintoshcomputers.Di?erentwordforms,pluralandsingular,arealsooftenbothpresent.InthisparticularsituationwiththeseMacintoshtags,the“relatedtags”sidebarofDeliciousinterlinksallthreeofthesecategoriesautomatically.Pluralvs.singularisoftenaproblem,asseeninthepopulartagsonFlickr,both“?ower”and“?owers”werelisted.
Thesesortsofproblemsarethereasonswhycontrolledvocabulariesareusedinmanysettings.Generally,anyoftheclassicproblemsthatcontrolledvocabularieshelpdealwithwillbepresentinthesesystemstovaryingdegrees.However,itislikelythatacontrolledvocabularywouldbeimpossibleinthecontextofsystemslikeDeliciousandFlickr.
3.5Strengths
Althoughafolksonomyisnotacontrolledvocabulary,andcertainlydoeshavelimitations,thereareimportantstrengthsthatareimportanttounderstandingtheappealandutilityofsuchsystems.
3.6Browsingvs.Finding
The?rstisserendipity.Whilethecontrolledvocabularyissuesdiscussedabovemayhamper?ndability,browsingthesystemanditsinterlinkedrelatedtagsetsiswonderfulfor?ndingthingsunexpectedlyinageneralarea.Inresearchingthispaper,exploringthebookmarkstaggedwith“folksonomy”onDelicious,thereweremanyrecentresourcesfromawidevarietyofauthorsandsitesthatIlikelywouldneverhavebeenexposedto.
Thereisafundamentaldi?erenceintheactivitiesofbrowsingto?ndin-terestingcontent,asopposedtodirectsearchingto?ndrelevantdocumentsinaquery.Itissimilartothedi?erencebetweenexploringaproblemspacetoformulatequestions,asopposedtoactuallylookingforanswerstospeci?callyformulatedquestions.Informationseekingbehaviorvariesbasedoncontext.WhileonecouldevaluateafolksonomyinasystemlikeDeliciousorFlickrbyusingspeci?cqueriesfromusers,andthenevaluatingwhichdocumentstaggedwithkeywordstheychoosearerelevanttothequery,thatwouldignorethebroadersetofbrowsingactivitiesthatthesystemseemstobestrongerin.Mea-suringtheutilityofthataspectwouldlikelyrequirequalitativeresearchinthe
6
formofinterviewsorethnographicstudyofusers,andisanareaoffurtherstudy.Itwouldalsorequirecomparisonsnottosearchbasedinformationretrievalsys-tems,buttobrowsingactivitiesusingothercategorizationandclassi?cationschemes.
3.7DesireLines
Perhapsthemostimportantstrengthofafolksonomyisthatitdirectlyre?ectsthevocabularyofusers.Inaninformationretrievalsystem,thereareatleasttwo,andpossiblymanymorevocabulariespresent(Buckland,1999).Thesecouldincludethatoftheuserofthesystem,thedesignerofthesystem,theauthorofthematerial,thecreatorsoftheclassi?cationscheme;translatingbetweenthesevocabulariesisoftenadi?cultandde?ningissueininformationsystems.Asdiscussedearlier,afolksonomyrepresentsafundamentalshiftinthatitisderivednotfromprofessionalsorcontentcreators,butfromtheusersofinformationanddocuments.Inthisway,itdirectlyre?ectstheirchoicesindiction,terminology,andprecision.
Someclassi?cationschemesaredisjointfromthevocabularyoftheusers.In“MetadatafortheMasses,”PeterMerholzarguesthatafolksonomycanbequiteusefulinthatitrevealsthedigitalequivalentof“desirelines”(Merholz,2004).Desirelinesarethefoot-wornpathsthatsometimesappearinalandscapeovertime.Merholznotes,“Asmartlandscapedesignerwillletwandererscreatepathsthroughuse,andthenpavetheemergingwalkways,ensuringoptimalutility.Ethnoclassi?cationsystemscansimilarlyemerge.’Onceyouhaveapreliminarysysteminplace,youcanusethemostcommontagstodevelopacontrolledvocabularythattrulyspeakstheusers’language.”
Merholzrecommendsusingafolksonomyasthestartofprofessionallyde-signedcontrolledvocabularies.Whilethismaynotbepracticalordesirableinmanysituations,thefundamentalpointisthatthevocabularyofusersmaysimplybetoodi?erentfromtheotherpartiestoadequately“pavethepaths”inadvance.Anotherimportantpointmaybethatthetermsuserswanttousemovetooquickly,orarequalitativelydi?erentthanauthorsorsystemsdesign-ers.
Merholz’sownpieceprovidesanexcellentexample.Merholzdoesnotusetheterm“folksonomy.”Hehaswrittenonhispersonalwebsitethatthetermisinaccurateduetoitsderivationfrom“taxonomy,”whichhearguestendtowardshierarchyandcontrol.(Merholz,2004)(SeealsoTaylor,2004,fordiscussionsofproblemsanddisputeswiththeterm“taxonomy.”)Merholzpreferstheterm“ethnoclassi?cation,”whichiswhatheusesinhisarticle,andthereisnomentionof“folksonomy”tobefound.Ethnoclassi?cationisalsoinaccurate,becauseasdiscussed,whatishappeningisquiteunlikeclassi?cationandfarmorelikecategorization.
DespiteMerholz’spersonalpreferenceasauthor,hispieceistaggedonDeli-ciouswithboth“ethnoclassi?cation”and“folksonomy,”aswellasvariousothertagsincluding“userexperience,”“tagging,”“taxonomy,”“metadata,”“social-software,”and“facets.”Thetagsre?ectnottheauthor’svocabulary,orany
7
particularclassi?cationorcategorizationsystem’svocabulary,butthelanguageandvocabularythatindividualuserschoosetodescribethearticlewith.
AlthoughtheDelicioustagsonMerholz’sarticleareonlyoneexample,afolksonomy,withitsuncontrollednatureandorganicgrowth,hasthecapabilitytoadaptveryquicklytouservocabularychangesandneeds.Thereisnosig-ni?cantcostforauserorforthesystemtoaddnewtermstothefolksonomy.Theproblemisthatwhilethedisparateuservocabulariesandtermsenablesomeveryinterestingbrowsingand?nding,thesheermultiplicityoftermsandvocabulariesmayoverwhelmthecontentwithnoisymetadatathatisnotusefulorrelevanttoauser.
4WhyFolksonomiesWork
Itisdi?culttode?neametricbywhichonecouldarguefolksonomiesareasuccessorfailure,butthedegreethatitdoesseemtobee?ectiveinthesesystemsasawayororganizinginformation,andthatalargegroupofpeopleareusingthesesystems,Iposit,isduetoafewimportantfactors.Theoverallcostsforusersofthesystemintermsoftimeande?ortarefarlowerthansystemsthatrelyoncomplexhierarchalclassi?cationandcategorizationschemes.Inadditiontothisstructuraldi?erence,thecontextoftheuseinthesesystemsisnotjustoneofpersonalorganization,butofcommunicationandsharing.Thenearinstantfeedbackinthesesystemsleadstoacommunicativenatureoftaguse.
4.1BarrierstoEntry,CognitiveCosts
Theconceptualshiftfromprofessional,designed,clearlyde?nedcategorizationandclassi?cationschemestoanad-hocsetofkeywordsenablesusersnotjustprofessionalswithoutanytrainingorpreviousknowledgetoparticipateinthesystemimmediately.Additionally,participatingisfareasierintermsoftime,e?ortandcognitivecosts.
StewartButter?eld,oneofthecreatorsofFlickr,arguesthatthedi?erenceincomplexitybetweenfolksonomiesandclassi?cationschemesisimportant:“Aside:Ithinkthelackofhierarchy,synonymcontrolandsemanticprecisionarepreciselywhyitworks.Freetypinglooseassociationsisjustaloteasierthanmakingadecisionaboutthedegreeofmatchtoapre-de?nedcategory(especiallyhierarchicalones).It’slike90%ofthevalueofaproper’taxonomybut10timessimpler.”(Butter?eld,2004)
ManyprofessionalswouldlikelyarguethatButter?eld’sassessmentof90%and“10timessimpler”isvastlyoverstated,hisfundamentalpointholdstrue:non-trivialandimportantmetadataarecapturedthroughthesefolksonomies.Thecomparisonsarealmostirrelevantasitwouldbeimpossibletogettheusersofthesesystemstouseacomplex,hierarchical,controlledvocabulary.Thebarriersaresimplytoohigh.Asystemthattriedtocapturethatfullvalue
8
wouldcosttoomuchinusertime,e?ort,andcognitivecost,andthushavelittlevalueinpractice.
4.2FeedbackandAsymmetricCommunication
JonUdell(2004)arguesthattheideaofabandoningtaxonomyinfavoroflistsofkeywordsisnotnew,andthatthefundamentaldi?erenceinthesesystemsisfeedback.
“Ofcourse,thatidea’sbeenaroundfordecades,sowhat’sspecialaboutFlickranddel.icio.us?Sometimesadi?erenceindegreebecomesadi?erenceinkind.Thedegreetowhichthesesystemsbindtheassignmentoftagstotheiruse-inatightfeedbackloop-isthatkindofdi?erence.
Feedbackisimmediate.Assoonasyouassignatagtoanitem,youseetheclusterofitemscarryingthesametag.Ifthat’snotwhatyouexpected,you’regivenincentivetochangethetagoraddanother.Ifyouritemsaren’tcon?-dentialandonline-onlyaccessissu?cient,thiscanbeagreatwaytomanagepersonalinformation.Buttherealpoweremergeswhenyouexpandthescopetoincludeallitems,fromallusers,thatmatchyourtag.Again,thatviewmightnotbewhatyouexpected.Inthatcase,youcanadapttothegroupnorm,keepyourtaginabidtoin?uencethegroupnorm,orboth.”(Udell,2004)
Thistightfeedbackloopleadstoaformofasymmetricalcommunicationbetweenusersthroughmetadata.Theusersofasystemarenegotiatingthemeaningofthetermsinthefolksonomy,whetherpurposefullyornot,throughtheirindividualchoicesoftagstodescribedocumentsforthemselves.
Therearetwomodelstodescribewhatishappeninghere:onethatfocusesonindividualincentives,andonethatfocusesoncommunityaspects.
4.3IndividualandCommunityAspects
BothDeliciousandFlickrareusedbyindividualstoorganizematerialswiththeirownvocabularyofterms.Individualshaveanincentivetotagtheirma-terialswithtermsthatwillhelpthemorganizetheircollectionsinawaythattheycan?ndtheseitemslater.Theorganizationalschemethatemergesforeachindividualre?ectstheirindividualinformationneeds.Thepopularityofthe“me”tagonFlickrperhapsbestre?ectsthisaspectofafolksonomy,aswellasthe“toread”tagonDelicious.Bothcanreallyonlybeunderstoodinthecontextofanindividualuser.
Conversely,bothDeliciousandFlickrareservicesdesignedtosharemateri-als.Theindividualorganizationalbehaviortakesplaceinapublicvirtualspaceonthesewebsites.Thereforethebehavioroftheuserscanalsobethoughtofasbeingin?uencedandrelatedtotheirrelationshiptotheotherindividualsusingtheservice,andspeci?cgroupsofuserswhotheysharetagusewith.Itisperhapshardertojustifythismodelsimplyfromexaminationofthetagsused,butthereisde?nitelyevidenceofcommunicationandperhapsevencommunityformationthroughmetadata,whichwillbediscussedlater.
9
Afolksonomylowersthebarrierstocooperation.Groupsofusersdonothave
toagreeonahierarchyoftagsordetailedtaxonomy,theyonlyneedtoagree,
inageneralsense,onthe“meaning”ofatagenoughtolabelsimilarmaterial
withtermsfortheretobecooperationandsharedvalue.Althoughthismay
requireachangeinvocabularyforsomeusers,itisneverforced,andasUdell
discussed,thetightfeedbackloopprovidesincentivesforthiscooperation.
Finally,thereisthecompulsiontoshareingeneralthatunderliesthese
systems.Theveryactofuserself-selectingwhattotagisimportant:thisisnot
justmaterialthatuserswantto?ndthemselveslater,butalsomaterialtheyare
sharingwithothers.Bothsystemshaveanexplicitkindofsocialnetworking
componentbuilt-in:Flickrallowsyoutospecifyotherusersascontacts,friends,
orfamilyandseeviewsofjusttheirmaterial;Deliciousallowsyouto“subscribe”
tootheruserslists.
Thesetwomodels,communityandindividualmotivations,arenotmutually
exclusive,anditislikelybotharenecessarytoexplainafolksonomyinthe
contextoftheseservices.Anareaoffurtherqualitativeanalysiscouldhelpto
determinehowmucheachofthesetheoriesappliestoactualuserbehavior.
4.4UnanticipatedUses
WhilethefolksonomiesthatdevelopedatFlickrandDelicioushaveade?nite
focusonsubjectcategorization,therearetagsbeingusedinsomeunexpected,
interestingwaysthatre?ectcommunicationandad-hocgroupformationfacili-
tatedthroughmetadata.
Flickr’s“sometaithurts”-for“sometaithurts”isacollectionofimagesre-
gardingFlickr,andpeopleusingFlickr.(http://www.?ickr.com/photos/tags/sometaithurts/)Theearliestimageisofsomeonediscussingsocialsoftware,andthensubsequent
userspostingscreenshotsofthatpicturewithinFlickr,andothersimilarlyself-
referentialimages.Thereferentialandmetanatureoftheimagescontinuesas
userstookpicturesofimagesonFlickr,etc.Althoughthisisaplayfulexample,
itisauseoftagsascommunicativetool.Onlybytaggingtheirphotographwith
“sometaithurts”couldauserofthesystemjointhephotographicconversation.
Conversely,theonlywaytofollowtheconversationwasthroughthesystems
automatedcollocatingofliketaggeditems.
AuseronFlickr,AndrewLowosky,beganpostingpicturesofdoorbellsin
Florence,alongwithabriefpieceof?ctionaboutthedoorbellinthedescription
ofthephotograph.Hedubbedthiscombinationofphotographandshortstory
“?icktion,”andtaggeditassuch.(Lowosky,2004.)Someotherusershavebeen
taggingphotographswith“?icktion”andwritingshort?ctiontoaccompanyit,
althoughasofNovember19,2004,therewereonlythreeotherusers.Although
small,thereisaquickformationofnewtermstodescribewhatisgoingon,and
othersadoptingthattermandtheactivityitdescribes.
ExaminingallphotosinFlickrtaggedwith“iraq”includesphotographs
Iraq,UStroopsinIraq,aswellasphotographsofwarprotests.Althoughthis
maynotbeacommunity,whatweareseeingisagroupofpeoplehelpingto
de?neatermwiththeirphotographsandmetadata.
10
Edited by Foxit ReaderCopyright(C) by Foxit Software Company,2005-2007For Evaluation Only.
55.1AreasForFurtherResearchQuantitativeTagAnalysis
Examiningthequantitativeaspectsoffolksonomiesisanareathatcouldyieldsomeinterestingdataonthemakeupanduseofthetermsusedtodescribeitems.Oneareatoexamineisthedistributionoftaguse:Ihypothesizethatitfollowsapowerlawscenario.Thatis,themostusedtagsaremorelikelytobeusedbyotheruserssincetheyaremorelikelytobeseen,andthustherewillbeafewtagsthatareusedbyasubstantialnumberofusers,thenanorderofmagnitudemoretagsthatareusedbyfewerusers,andanotherorderofmagnitudemoreusedbyonlyahandfulofusers.Examiningthissortofdistributionoftagusecouldgiveabetterindicationofwhetherafolksonomyconvergesontermsandfosterconsensus,orifastheuserbasedgrowsthevocabularygrowsatamoreevenrate,andthedistributionofterms?attens,perhapsindicatinglessagreement.
5.2QualitativeUserAnalysis
Examininguserbehaviorthroughethnographicobservationorinterviewtoun-derstandusermotivationsandcognitiveprocessesintaggingitemswouldclarifywhatfactorsdirectlyin?uencetheformationofafolksonomy,andhowindividualincentivesandgroupcommunicationmotivationsin?uenceuseofthesystem.Althoughitseemsthatsomeusersareintendingtofacilitatecommunicationthroughtaguse,especiallyintheunintendeduses,interviewscouldmakethispointexplicit.Interviewscouldalsoelucidatetheconsciousintentionsofusersin“normal”useofthesystem,whichismuchhardertoobservesimplyfromthedocumentsandtagsthemselves.Otherbehaviorthatwouldbehelpfultoobserveisthefrequencywithwhichusersmodifyorchangetheirtags,orfuturetaggingbehaviorbasedontheimplicitfeedbackfromthesystemintheformofwhatotherdocumentsaretaggedwithaterm.
5.3Applicabilitytoothersystems
DeliciousandFlickrarelargewebservicesdesignedtoorganizeandsharedigi-talworks.Theapplicabilityofusergeneratedfree-formtaggingasanorganiza-tionalconstructinothercontextsbearsfurtherinvestigation.Oneinterestingareawouldbetosystemswherethereisalreadyanexistingsocialnetworkandexaminehowthisrelatedtothesystem.Acorporateintranet,orasysteminanacademicsettingusedbyadepartment’sfacultyandstudents.
Theuseofafolksonomytosupplementexistingclassi?cationschemesandprovideadditionalaccesstomaterialsbyencouragingandleveragingexplicitusermetadatacontributionsisapossibleareaforresearchandfurtherdevel-opmentininformationretrievalsystems.Ifinformationretrievalsystemsbegintoincorporateuser-centeredinformationmanagementtools,theorganizational
11
Edited by Foxit ReaderCopyright(C) by Foxit Software Company,2005-2007For Evaluation Only.
schemesdevelopedbytheusershavethepossibilitytobeofgreatinterestto
otherusersandimprovethesystems.
6ConclusionAfolksonomyrepresentssimultaneouslysomeofthebestandworstintheorga-
nizationofinformation.Itsuncontrollednatureisfundamentallychaotic,suf-
fersfromproblemsofimprecisionandambiguitythatwelldevelopedcontrolled
vocabulariesandnameauthoritiese?ectivelyameliorate.Conversely,systems
employingfree-formtaggingthatareencouraginguserstoorganizeinformation
intheirownwaysaresupremelyresponsivetouserneedsandvocabularies,and
involvetheusersofinformationactivelyintheorganizationalsystem.Overall,
transformingthecreationofexplicitmetadataforresourcesfromanisolated,
professionalactivityintoashared,communicativeactivitybyusersisanim-
portantdevelopmentthatshouldbeexploredandconsideredforfuturesystems
development.
7References
1.Blood,Rebecca.“Weblogs:AHistoryandPerspective”,Rebecca’sPocket.
September7,2000.[/essays/webloghistory.html]
2.Butter?eld,Stewart.“Sylloge.”August4,2004.[/personal/2004/08/folksonomysocial-classi?cation-great.html]
3.Buckland,Michael.“VocabularyasaCentralConceptinLibraryandIn-
formationScience.”DigitalLibraries:InterdisciplinaryConcepts,Chal-
lenges,andOpportunities.ProceedingsoftheThirdInternationalConfer-
enceonConceptionsofLibraryandInformationScience(CoLIS3,Dubrovnik,
Croatia,23-26May1999.Ed.byT.Arpanacetal.Zagreb:Lokve,pp?3-12.[http://www.sims.berkeley.edu/buckland/colisvoc.htm]
4.Greenberg,Jane,MariaCristinaPattuelli,BijanParsiaandW.Davenport
Robertson.“Author-generatedDublinCoreMetadataforWebResources:
ABaselineStudyinanOrganization.”JournalofDigitalInformation,
Volume2Issue2ArticleNo.78,November2001.[http://jodi.ecs.soton.ac.uk/Articles/v02/i02/Greenberg
5.Jacob,ElinK.“Classi?cationandcategorization:adi?erencethatmakesa
di?erence.”LibraryTrends.Winter,2004.[http://www.?ndarticles.com/p/articles/mim1387/is52/ai
6.Lieberman,Henry.“InterfacesThatGiveandTakeAdvice.”FromHuman-
ComputerInteractionintheNewMillenium.JohnM.Carroll,Ed.Boston:
AddisonWesley,2002.
7.Losowsky,Andrew.“Thedoorbellsof?orenceproject.”ThePrandial
Post.September21,2004.[/archives/200412
8.Merholz,Peter.“Ethnoclassi?cationandvernacularvocabularies.”Au-
gust30,2004.[/archives/000387.html]
9.Merholz,Peter.“MetadatafortheMasses.”October19,2004.[/publicatio
10.Page,L.;Brin,S.;Motwani,R.;andWinograd,T.1998.ThePageR-
ankcitationranking:Bringingordertotheweb.[http://wwwdb.stan-
ford.edu/backrub/pageranksub.ps]
11.Schachter,Joshua.Del.icio.usAboutPage.[http://del.icio.us/doc/about].
2004.
12.Smith,Gene.“Atomiq:Folksonomy:socialclassi?cation.”Aug3,2004[http://atomiq.org/archives/2004/08/folksonomyclassi?cation.html]
13.Taylor,ArleneG.TheOrganizationofInformation.Westport:Libraries
Unlimited2004.
14.Udell,Jon.“Collaborativeknowledgegardening.”InfoWorld.August20,
2004.[/article/04/08/20/34OPstrategic1.html
13