247K - views

What Trends in Chinese Social Media Louis Yu Social Computing Lab HP Labs Palo A

yuhpcom Sitaram Asur Social Computing Lab HP Labs Palo Alto California USA sitaramasurhpcom Bernardo A Huberman Social Computing Lab HP Labs Palo Alto California USA bernardohubermanhpcom ABSTRACT There has been a tremendous rise in the growth of onl

Download Pdf

What Trends in Chinese Social Media Louis Yu Social Computing Lab HP Labs Palo A




Download Pdf - The PPT/PDF document "What Trends in Chinese Social Media Loui..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.



Presentation on theme: "What Trends in Chinese Social Media Louis Yu Social Computing Lab HP Labs Palo A"— Presentation transcript:

Therehavebeenmanyexperimentsconductedforstudy-ingthestructureofonlinesocialnetworks.Inonecompre-hensivestudy,Misloveetal.[23]havepresentedalarge-scalemeasurementstudyofonlinesocialnetworkssuchasOrkut,YouTube,andFlickr.Theirresultsshowthatonlinesocialnetworksfollowthepower-lawinthein-degreeandout-degreedistributionsofusernodes.Inotherwork,Ku-maretal.[18]examinethelinkingstructureofFlickrandYahoo!360andreportsimilar ndings.2.2SocialInuenceStudiesFormanyyearsthestructureofvariousoinesocialnet-workshasbeenstudiedbysociologists(see[15][23][6]forsurveys).ResearchershavealsoanalyzedthestructureofvariousChineseoinesocialnetworks[4][25][12][5][7].Insocialnetworkanalysis,socialin uencereferstotheconceptofpeoplemodifyingtheirbehaviortobringthemclosertothebehavioroftheirfriends.Socialin uencehasbeenstudiedinavastarrayofsocialnetworksinvolvingvariousfocisuchasinterestsandper-sonalhabits[22][13][26].Xuetal.[30]havelookedattheadaptationofaggressivebehaviorsinasocialnetworkofkindergardenchildreninChina.Asamethodofcontrollingaggression,teachersinChinatendtoputaggressivechil-dreninapeergroupwithnon-aggressivechildren.Xuetal.[30]havefoundthatovertimefriendshipscanbeformedbetweenaggressivechildrenandnon-agressivechildren.Fortheaggressivechildrenwhoaregroupmembers,intra-groupfriendshipsmoderatedtheiraggressivebehavior.Agarwaletal.[1]haveexaminedtheproblemofidenti-fyingin uentialbloggersintheblogosphere.Theydiscov-eredthatthemostin uentialbloggerswerenotnecessarilythemostactive.Backstrometal.[3]haveexaminedthecharacteristicsofmembershipclosureinLiveJournal.andCrandalletal.[11],theadaptationofin uencesbetweeneditorsofWikipediaarticles.OnTwitter,Chaetal.[8]haveperformedacomparisonofthreedi erentmeasuresofin uence-indegree,retweetsandusermentions.Basedonthis,theyhypothesizedthatthenumberoffollowersmaynotagoodmeasureofin uence.ThiswascorroboratedbyRomeroandothers[24]whopresentedanovelin uencemea-surethattakesintoaccountthepassivityoftheaudienceinthesocialnetwork.TheymeasuredretweetsonTwitterandfoundthatpassivitywasamajorfactorwhenitcametofor-warding.Theyalsodemonstratedwithempiricalevidencethatthenumberoffollowersisapoormeasureofin uence.Thereareonlyafewstudiesofsocialin uenceinChi-neseonlinesocialnetworks.Jin[16]hasstudiedtheChineseonlineBulletinBoardSystems(BBS),andprovidedobser-vationsonthestructureandinterfaceofChineseBBSandthebehavioralpatternsofitsusers.Xin[29]hasconductedasurveyonBBS'sin uenceontheUniversitystudentsinChinaandtheirbehavioronChineseBBS.Yuetal.[32]haslookedattheadaptationofinterestssuchasbooks,movies,music,eventsanddiscussiongroupsonDouban,anonlinesocialnetworkfrequentlyusedbytheyouthinChina.Doubanprovidesuserswithreviewandrecommendationser-vicesformovies,books,musicandevents.ItisalsothelargestonlinemediadatabaseandoneofthelargestonlinecommunitiesinChina.2.3TrendsonTwitterTherearevariousstudiesontrendsonTwitter[14][19][21][28].Recently,Asurandothers[2]haveexaminedthegrowthandpersistenceoftrendingtopicsonTwitter.Theydiscoveredthattraditionalmediasourcesareimportantincausingtrendsontwitter.Manyofthetopretweetedarti-clesthatformedtrendsonTwitterwerefoundtoarisefromnewssourcessuchastheNewYorkTimes.Inthiswork,weevaluatehowthetrendingtopicsinChinarelatetothenewsmedia.2.4TheInternetinChinaThedevelopmentoftheInternetindustryinChinaoverthepastdecadehasbeenimpressive.AccordingtoasurveyfromtheChinaInternetNetworkInformationCenter(CN-NIC),byJuly2008,thenumberofInternetusersinChinahasreached253million,surpassingtheU.S.astheworld'slargestInternetmarket[9].Furthermore,thenumberofIn-ternetusersinChinaasof2010wasreportedtobe420million.Despitethis,thefractionalInternetpenetrationrateinChinaisstilllow.The2010surveybyCNNIContheIn-ternetdevelopmentinChina[10]reportsthattheInternetpenetrationrateintheruralareasofChinaisonaverage5:1%.Incontrast,theInternetpenetrationrateintheur-bancitiesofChinaisonaverage21:6%.InmetropolitancitiessuchasBeijingandShanghai,theInternetpenetra-tionratehasreachedover45%,withBeijingbeing46:4%andShanghaibeing45:8%[10].AccordingtothesurveybyCNNICin2010[9],China'scyberspaceisdominatedbyurbanstudentsbetweentheageof18{30(seeFigure1andFigure2,takenfrom[9]). Figure1:AgeDistributionofInternetUsersinChinaTheGovernmentplaysanimportantroleinfosteringtheadvanceoftheInternetindustryinChina.Tai[31]pointsoutthefourmajorstagesofInternetdevelopmentinChina,\witheachperiodre ectingasubstantialchangenotonlyintechnologicalprogressandapplication,butalsointheGovernment'sapproachtoandapparentperceptionoftheInternet."1.The rstphasewasbetween1986{1992,whenInternetapplicationswerelimitedtotheuseofemailsamongahandfulofcomputerresearchlabsinChina.2.Thesecondphasewasbetween1992{1995,theChi-neseGovernmentproposedseverallargescalenetworkprojectsandbuiltanationalinformationnetworkin-frastructure. Figure2:TheOccupationDistributionofInternetUsersinChina3.Thethirdphasewasbetween1995{1997.TheChineseGovernmentsteppedupitse ortinbuildingthein-formationnetworkinfrastructure,hopingthattheITindustrywouldyieldsigni cantbene tstothenation'seconomy.Meanwhile,theGovernmentstartedtoim-plementavarietyoftechnologicalandpolicycontrolmechanismstoregulatethesafe owoftheinformationontheInternet.4.Thefourthphasestartedfrom1998andcontinuestothepresent,duringwhichtimetheInternethasbecomeapowerfulmediumintheChinesesociety.AccordingtoTheInternetinChina1releasedbytheIn-formationOceoftheStateCouncilofChina:TheChinesegovernmentattachesgreatimpor-tancetoprotectingthesafe owofInternetin-formation,activelyguidespeopletomanageweb-sitesinaccordancewiththelawandusetheIn-ternetinawholesomeandcorrectway.2.5ChineseOnlineSocialNetworksOnlinesocialnetworksareamajorpartoftheChineseInternetculture[16].Netizens2inChinaorganizethem-selvesusingforums,discussiongroups,blogs,andsocialnet-workingplatformstoengageinactivitiessuchasexchangingviewpointsandsharinginformation[16].AccordingtoTheInternetinChina:Vigorousonlineideasexchangeisamajorchar-acteristicofChina'sInternetdevelopment,and 1\TheInternetinChina"bytheInformationOceoftheStateCouncilofthePeople'sRepublicofChinaisavailableathttp://www.scio.gov.cn/zxbd/wz/201006/t667385.htm2Anetizenisapersonactivelyinvolvedinonlinecommuni-ties[27].thehugequantityofBBSpostsandblogar-ticlesisfarbeyondthatofanyothercountry.China'swebsitesattachgreatimportancetopro-vidingnetizenswithopinionexpressionservices,withover80%ofthemprovidingelectronicbul-letinservice.InChina,thereareoveramillionBBSsandsome220millionbloggers.Accordingtoasamplesurvey,eachdaypeoplepostoverthreemillionmessagesviaBBS,newscommen-tarysites,blogs,etc.,andover66%ofChinesenetizensfrequentlyplacepostingstodiscussvar-ioustopics,andtofullyexpresstheiropinionsandrepresenttheirinterests.Thenewapplica-tionsandservicesontheInternethaveprovidedabroaderscopeforpeopletoexpresstheiropin-ions.Thenewlyemergingonlineservices,includ-ingblog,microblog,videosharingandsocialnet-workingwebsitesaredevelopingrapidlyinChinaandprovidegreaterconvenienceforChinesecit-izenstocommunicateonline.Activelypartici-patinginonlineinformationcommunicationandcontentcreation,netizenshavegreatlyenrichedInternetinformationandcontent.3.SINAWEIBOFromtheabovemotivation,wethinkitisinterestingtolookathowtrendsstartandevolveinvariousChineseon-linesocialnetworksandtoanalyzethecharacteristicsoftrend-settersdeterminingiftheyrepresentGovernmentor-ganizations,commercialorganizations,themedia,orindi-viduals.Wechoosetoanalyzethecharacteristicsoftrendsandtrend-settersonSinaWeibo.SinaWeibowaslaunchedbytheSinacorporation,China'sbiggestwebportal,inAu-gust2009.OnJuly2009,theChineseGovernmentblockedtheaccesstoTwitterandFanfou,thethenleadingTwit-terclone,inChina.InternetcompaniessuchasSinaandTencentstartedo eringmicroblogservicestotheirusersinmainlandChina.AccordingtotheSinacorporationannualreport3,theWeibomicroblognowhasmorethan50millionactiveusersperday,and10millionnewlyregistereduserspermonth.WhilebothTwitterandSinaWeiboenableuserstopostmessagesofupto140characters,therearesomedi erencesintermsofthefunctionalitieso ered.Wegiveabriefintro-ductionofSinaWeibo'sinterfaceandfunctionalities.3.1UserProlesAuserpro leonSinaWeibodisplaystheuser'sname,abriefdescriptionoftheuser,thenumberoffollowersandfolloweestheuserhas,andthenumberoftweetstheusermade.Auserpro lealsodisplaystheuser'srecenttweetsandretweets.SimilartoTwitter,therearetwotypesofuseraccountsonSinaWeibo,regularuseraccountsandveri eduseraccounts.Averi eduseraccounttypicallyrepresentsawellknownpublic gureororganizationinChina.Sinahasreportedintheannualreportthatithasmorethan60,000veri edaccountsconsistingofcelebrities,sportsstars,wellknownorganizations(bothGovernmentandcommercial)andother 3TheSinacorporationannualreport2011isavail-able(inChinese)athttp://tech.sina.com.cn/i/2010-11-16/10314870771.shtml VIPs.3.2TheContentofTweetsonSinaWeiboThereisanimportantdi erenceinthecontentoftweetsbetweenSinaWeiboandTwitter.WhileTwitteruserscanposttweetsconsistingoftextandlinks,SinaWeibouserscanpostmessagescontainingtext,pictures,videosandlinks.Figure3illustratessomemessageswithembeddedpicturesandvideosonSinaWeibo. Figure3:AnExampleofEmbeddedVideosandPic-tures(TranslationsoftheTweetsOmitted)3.3RetweetsandCommentsTwitteruserscanaddresstweetstootherusersandcanmentionothersintheirtweets[13].AcommonpracticeonTwitteris\retweeting",orrebroadcastingsomeoneelse'smessagestoone'sfollowers.TheequivalentofaretweetonSinaWeiboisinsteadshownastwotwoamalgamateden-tries:theoriginalentryandthecurrentuser'sactualentrywhichisacommentaryontheoriginalentry(seeFigure4).SinaWeiboalsohasafunctionalityabsentfromTwitter:thecomment.WhenaWeibousermakesacomment,itisnotrebroadcastedtotheuser'sfollowers.Instead,itcanonlybeaccessedundertheoriginalmessage.Figure4illustratestwoexampletweetsonSinaWeibo.The rstisanoriginaltweetmadebyauser,wecanseethattheretweetingandcommentingbuttonsarelistedunderthetweet.Thesecondisaretweet,wecanseethattheoriginalmessageisretweeted62timesandcommented10timesbyotherusers.3.4TrendingkeywordsSinaWeiboo ersalistof50keywordsthatappearedmostfrequentlyinusers'tweetsoverthepasthour.Theyarerankedaccordingtothefrequencyofappearances.Figure5 Figure4:AnExampleofCommentsandRetweets(TranslationsoftheTweetsOmitted)illustratesthelistofhourlytrendingkeywords(withtransla-tions).ThisissimilartoTwitter,whichalsopresentsacon-stantlyupdatedlistoftrendingtopics,whicharekeywordsthataremostfrequentlyusedintweetsoverthatperiod. Figure5:TheListofHourlyTrendingKeywords(withTranslations)Wemonitoredthelistofhourlytrendingkeywordseveryhourfor30daysandretrievedeverynewkeywordsappearedinthelist.Weretrievedintotal4411newtrendingkeywordsoverthe30daysobservationperiod.TocomparewithTwitter,weobtained16.32milliontweetson3361di erenttrendingtopicsover40daysusingtheTwitterSearchAPI.4.EXPERIMENTSANDRESULTSFirst,wecalculatedthedistributionsforthenumberoftweetsandtopicsinourdataset.Figure6a)illustrates Figure6:TheDistributionfortheNumberofTweetsandtheNumberofTopicsTable3:Pro leInformationforTop20RetweetedUsers Images(%) Videos(%) Links(%) Followees Followers Tweets 1 70% 0% 32% 673 461398 719 2 57% 71% 0% 715 300358 2508 3 21% 0% 17% 67 597063 2600 4 30% 0% 0% 20 245026 518 5 20% 0% 0% 81 1884896 4261 6 8% 0% 0% 650 3536888 10598 7 15% 11% 0% 12 625117 804 8 46% 0% 0% 368 2338610 3605 9 0% 22% 0% 79 405847 716 10 5% 1% 0% 11 2411888 17818 11 14% 4% 0% 634 1551438 4899 12 44% 16% 0% 352 6107858 1850 13 0% 50% 0% 1136 590099 2041 14 6% 7% 0% 303 1210833 11411 15 10% 0% 1% 555 1220027 4249 16 45% 0% 0% 13 615461 1254 17 5% 40% 0% 12 496171 571 18 30% 0% 0% 60 901612 3506 19 15% 3% 0% 9 763264 2718 20 25% 0% 0% 4 853877 2362 Figure7:AnIllustrationofanUserAccountonSinaWeibodominatedbypopularnewssourcessuchasCNN,theNewYorkTimesandESPN.Alargepercentageofthetopicsthattrendedaccordinglydealtwitheventsinthenews.Thisin-dicatesthatTwitterusersaremoreattunedtonewseventsthanSinaWeibousersandamplifysucharticlesthroughthemediumofTwitter.Ontheotherhand,theconsistenttrend-settersonSinaWeiboarenotmediaorganizations.Insteadtheyareunveri edaccountsactingasdiscussionfo-rumsandaplatformforuserstosharefunnypictures,jokes,andstories.Weobservethatnoneoftheunveri edaccountsinTable1arepersonalaccounts,withonly1outof4veri edaccounts(inthetop20)belongingtoamediaorganization.Thisrepresentsanimportantcontrastintheuseofthesemedia,withChineseusersbeingmoreinclinedtoshareandpropagatetrivialcontentthantheTwitterusers.4.1.2RetweetsWhenweconsidertheratioofretweets,weonceagainobserveastrongcontrastwithTwitter.ThenumberofretweetsthatauthorsgetonSinaWeiboareseveralordersofmagnitudegreaterthantheretweetsforthetrendingtop-icsonTwitter,althoughtheycontributetofewertopics.Thisimpliesthatthetopicsaretrendingmainlybecauseofsomecontentthathasbeenretweetedmanytimes.TheTweetscolumninTable1givestheuniquetweetsthathavebeenretweeted.Wecanobservethattherateatwhichtheyhavebeenretweetedisphenomenal.Forexample,thetopretweeteduserposted37tweetswhichweretotallyretweeted1194999times.Theoverallretweetpercentagewasaround62%forthetrendingtopics.Incontrast,forTwittertrends,theretweetsformonly31%oftheoveralltweetsforthe Author Followees Followers vovo panico 1069 154589 cnnbrk 41 4380908 keshasuja 0 88 LadyGonga 37 136433 BreakingNews 382 2570662 MLB 18829 1237615 nytimes 465 3250977 HerbertFromFG 763 23318 espn 286 1326168 globovision 3582 753440 hungtonpost 4684 1042330 skynewsbreak 5 198349 el pais 46226 572260 stcom 12 59763 la patilla 51 306965 reuters 603 724204 WashingtonPost 284 458721 bbcworld 20 796009 CBSnews 122 1716649 TelegraphNews 238 38599 Table4:Follower/FolloweerelationshipsforTopRetweetedTwitterUserstrendingtopics.WhileretweetsdocontributetomakingatopictrendonTwitter,theire ectisnotsolarge.4.1.3EmbeddedImages,VideosandLinksinTweetsForthelistoftrend-settersonSinaWeibo(Table1),wealsoexaminethecontentoftheirtweetsthatappearedinthetrendingtopics.Table3illustratesthepercentageoftheseusers'tweetswhichincludedimages,videos,andlinks.Weobservethatalargepercentageofthetweetsoftheseusersincludeanembeddedimage,andmanytweetsincludedanembeddedvideooralink.Ontheotherhand,Twitteruserspostlinksinonly17.6%ofthetweetsontrendingtopics.Thisisagaindemonstrativeofthetypeofcontentthatissharedinthesetwosocialmediaservices.4.1.4FollowerRelationshipsForeachofthein uentialauthors,welookedatthenum-beroffollowersandfolloweestheyhave,andthetotalnum-bertweetstheyhavemadesincetheiraccountsareactivated(Table3).Wediscoveredthatmostofthein uentialauthorshavemorefollowersthanfollowees.Wehypothesizethatthesein uentialauthorsdonotactivelyseekoutaccountstofollow.Itistheircontentwhichattractsotheruserstofollowthem.Interestingly,wefoundthistobetrueonTwit-teraswell(showninTable4),withthetopretweetedusershavingveryskewedfollower/followeeratios.4.1.5VeriedAccountsWhenweconsideredthetop100trend-setters,wefoundthatonly23wereveri edaccounts.Thedescriptionsofthese23areshowninTable5.Weobservedtheveri edaccountsareofcelebrities,newspapers,magazinesandafewothermediasources.4.2RandomProleAnalysisFromtheaboveanalyses,weobservethattheSinaWeibouseraccountswhosetweetsareretweetedfrequentlyandap-pearinmultipletrendingtopicsovertimearemostlyestab- Rank ID Description 1 1757128873 FashionWebMagazine 2 1643830957 FashionBrand 2 1670645393 TravelWebMagazine 12 1195230310 Celebrity 21 1740006601 Celebrity 25 1730380283 GameDiscussionForum 26 1760945071 ChineseGroupon 42 1322920531 Celebrity 43 1771665380 RecordLabel 46 1266321801 Celebrity 48 1883881851 Organization(NBAChina) 58 1698229264 MusicWebMagazine 62 1642591402 SinaEntertainment 70 1743374541 PicturesDiscussionForum 71 1618051664 SinaNews 74 1653689003 Newspaper 75 1640601392 SinaVideo 82 1195031270 Celebrity 83 1835254597 MusicWebMagazine 84 1830442653 MusicWebMagazine 95 1765148101 SinaFashion 96 1258256457 Celebrity 100 1596329427 Celebrity Table5:Veri edAccountsAmongTop100Trend-setterstrac.ThisisaverylowpercentagewhencomparedtoSinaWeibo.5.CONCLUSIONANDFUTUREWORKWeanalyzedthetweetsthatareresponsibleforcausingtrendingtopicsonSinaWeiboandtheusersthatcreatedthesetweets.Weobservedthattherearevastdi erencesbetweenthecontentthatissharedonSinaWeibothanthatofTwitter.PeopletendtouseSinaWeibotosharejokes,imagesandvideosandasigni cantlylargepercentageofpostsareretweets.Thetrendsthatareformedarealmostentirelyduetotherepeatedretweetsofsuchmediacontent.Incontrast,weobservedonTwitterthattrendingtopicsaremainlycausedbysourcesofmedia[2].Asfuturework,weplantoconductfurthertemporalanal-ysisregardingtheevolutionoftrendsonSinaWeibo.Wealsoplantoidentifythepatternsofretweetsbetweenfollow-ersofin uentialauthors.6.REFERENCES[1]N.Agarwal,H.Liu,L.Tang,andP.S.Yu.IdentifyingtheIn uentialBloggersinaCommunity.WSDM'08,2008.[2]S.Asur,B.A.Huberman,G.Szabo,andC.Wang.Trendsinsocialmedia-persistenceanddecay.In5thInternationalAAAIConferenceonWeblogsandSocialMedia,2011.[3]L.Backstrom,D.Huttenlocher,J.Kleinberg,andX.Lan.Groupformationinlargesocialnetworks:membership,growth,andevolution.InProceedingsofthe12thInternationalConferenceonKnowledgeDiscoveryandDataMining,pages44{54.ACM,2006.[4]Y.Bian.Bringingstrongtiesbackin:indirectties,networkbridges,andjobsearchesinchina.AmericanSociologicalReview,62(3):366{385,1997.[5]Y.Bian,R.Breiger,D.Davis,andJ.Galaskiewicz.Occupation,class,andsocialnetworksinurbanchina.SocialForces,83(4):1443{1468,2005.[6]M.Buchanan.Nexus:SmallWorldsandtheGroundbreakingTheoryofNetworks.W.W.Norton&Company,May2003.[7]P.J.Carrington,J.Scott,andS.Wasserman,editors.ModelsandMethodsinSocialNetworkAnalysis.CambrigeUniversityPress,2005.[8]M.Cha,H.Haddadi,F.Benevenuto,andK.P.Gummadi.MeasuringUserIn uenceinTwitter:TheMillionFollowerFallacy.InFourthInternationalAAAIConferenceonWeblogsandSocialMedia,May2010.[9]CNNIC.The21ststatisticsreportontheinternetdevelopmentinchina(inchinese),2010.[10]CNNIC.Surveyreportoninternetdevelopmentinruralchina(inchinese),2010.[11]D.Crandall,D.Cosley,D.Huttenlocher,J.Kleinberg,andS.Suri.Feedbacke ectsbetweensimilarityandsocialin uenceinonlinecommunities.InProceedingsofthe14thACMSIGKDDinternationalconferenceonKnowledgediscoveryanddatamining,pages160{168.ACM,2008.[12]J.-L.Farh,A.S.Tsui,K.Xin,andB.-S.Cheng.Thein uenceofrelationaldemographyandguanxi:theChinesecase.OrganizationScience,9(4):471{488,1998.[13]A.Goyal,F.Bonchi,andL.V.S.Lakshmanan.Learningin uenceprobabilitiesinsocialnetworks.InWebSearchandDataMining,pages241{250,2010.[14]B.A.Huberman,D.M.Romero,andF.Wu.Socialnetworksthatmatter:Twitterunderthemicroscope.ComputingResearchRepository,2008.[15]M.JamaliandH.Abolhassani.Di erentaspectsofsocialnetworkanalysis.InProceedingsofthe2006IEEE/WIC/ACMInternationalConferenceonWebIntelligence,pages66{72,2006.[16]L.Jin.ChineseoutlineBBSsphere:whatBBShasbroughttoChina.Master'sthesis,MassachusettsInstituteofTechnology,April2009.[17]D.Kempe,J.Kleinberg,andE.Tardos.In uentialnodesinadi usionmodelforsocialnetworks.InProceedingsof32ndInternationalColloquiumonAutomata,LanguagesandProgramming,pages1127{1138.SpringerVerlag,2005.[18]R.Kumar,J.Novak,andA.Tomkins.Structureandevolutionofonlinesocialnetworks.InProceedingsofthe12thACMSIGKDDInternationalConferenceonKnowledgeDiscoveryandDataMining,pages611{617.ACM,2006.[19]H.Kwak,C.Lee,H.Park,andS.Moon.Whatistwitter,asocialnetworkoranewsmedia?InProceedingsofthe19thinternationalconferenceonWorldwideweb,WWW'10,pages591{600,2010.[20]J.Leskovec,A.Singh,andJ.Kleinberg.Patternsofin uenceinarecommendationnetwork.InProceedingsofPaci c-AsiaConferenceonKnowledgeDiscoveryandDataMining,pages380{389.Springer-Verlag,2005.