Escolar Documentos
Profissional Documentos
Cultura Documentos
APictureisWorth32.33Words:ImportanceofAnalyzingImagesonOnlineSocialMedia|PreCog'sBlog
APictureisWorth32.33Words:Importanceof
AnalyzingImagesonOnlineSocialMedia
PostedonAugust1,2016byPK
Doyourememberthelasttimeyourushedorsawanyonerushtogetanautographofafamouspersonality?No,
right?Becausethosedaysarelonggone.Todaysgenerationbelievesintakingaselfieinstead.Andwhynot,digital
mediaisforever,oratleast,itcaneasilyoutliveapieceofpaperwithanautograph!Thereisanexplosionofdata
thatisgeneratedontheOnlineSocialMedia(OSM),wesee422,340tweetsonTwitter,3.3millionupdateson
Facebook,55,555picturesuploadedonInstagrameverysecond[1].Intherecentpast,withtheupdates,large
fractionofitisimages/picturesoneanalysisshowsthat1.8billionphotosaresharedonFacebook,Instagram,
Flickr,Snapchat,andWhatsAppeveryday[2].Itisalsofoundthatupdateswithimagesincreasetheengagementof
theposts,like[3]shows18%moreclicks,89%morefavorites,and150%moreretweetswhenthetweethasan
imagecomparedtoonlytextupdates.Anotherarticlereports93%ofthemostengagingpostsonFacebookhavean
image[4].ResearchersarealsostudyingwhatmakesanimagepopularonnetworkslikeFlickr[5].
Inlastfewyearstherehavebeenmanyacademicpapers,technologiesinrealworldalllookingatthisgrowthof
contentandanalyzingthemweseemostofthemanalyzingonlythetextualpartofthecontent.Hereisanon
comprehensivelistofpublicationsinsomeofthetoptierconferencesinthisspaceallofthesepaperslookat
contentgeneratedinEnglish[620].Someresearchersarealsolookingatstudyingthesentimentandtextual
characteristicsofnonEnglishcontentonOSM[2127].Languagesinclude,Farsi,andHindi.
IhavebeencuriousforalittlewhilenowaboutnontextualcontentonOSMsomeofmyrecentinteresthasbeento
lookatimagesandvideosonOSM.IrecentlyhadmystudentSonalGoelinvestigateimagesonOSM,she
completedherMastersthesisImageSearchforImprovedLawandOrder:Search,Analyse,Predictimagespread
onTwitterwhereshepredictedtheviralityofimagesonOSMusingtweetsfrommultipleevents.PrateekDewan,my
Ph.D.studentandIhavebeenplayingaroundthebroadertopicofimagesandOSM.Webelievethattheinferences
thatwedrawfromtextualanalysiscanbedifferentfromtheanalysisdonewithimagesfromthesameposts.For
example,textualanalysisdoneinHurricaneSandy[28]andBostonMarathon[29]couldhaveclassifiedtheposts
withimages(alongwithtext)tobelegitimate,whereas,ifweanalyzetheimagesitselfitmaybefake.Belowisafake
imagewhichwentviralduringSandy,buttextualanalysisforthepostswiththeseimagescouldhaveleanedtowards
crediblecontent.
http://precog.iiitd.edu.in/blog/2016/08/imagesononlinesocialmedia/
1/5
12/3/2016
APictureisWorth32.33Words:ImportanceofAnalyzingImagesonOnlineSocialMedia|PreCog'sBlog
SentimentanalysisoftheOSMcontentisusedtomakedecisionsonthepulseofcitizens,customers,etc.
Sometimesthesentimentofthetextualcontentisverydifferentfromtheimagespostedwiththetext.Belowimage
waspostedwiththecontentThankyouPiersMorganforspeakingtruth.#PrayForParis#MuslimsStandWithParis
[30]Textanalysiswillgivepositive/neutralsentiment,whilethecontentfromtheimageattachedwiththepostis
negative.Wefoundotherexamplestosubstantiatethispoint,postbeingnegativeandimagebeingmorepositive[33,
34]andpostbeingpositiveandimagebeingmorenegative[35].
Justtotestourhypothesisofhowmuchinformationisspreadthroughimages,weanalyzedsomeeventsforwhich
wehavebeencollectingdata.Belowisthetablewhichshowsdatafor9eventsconsistentlyweseethatonaverage
about2025%ofthecontenthasonlyimageswithouttext.Inmostoftheanalysisthatisdonenowwithtextual
contentwillmissthisinformation.Inoneoftheeventthatweareanalyzingnow,wewereabletoextracttextfrom
8,200imagestheseimageswerepostedonOSMwithnotext.Tounderstandtheamountoftextthatareshared
throughimages,wegotimagesannotatedandusingTesseractOCR[31],wewereabletoget1,030,471wordsfrom
31,869images.
ColumnwithtextreferstothenumberofpostscontainingthemessagefieldasreturnedbytheGraphAPI.This
fieldcontainsthestatus/textmessagepostedbytheuser.Thewithimagecolumnrepresentsthenumberofposts
wherethetypeofpostisphoto.Facebookautomaticallydeterminesthistypewhileauseriscomposingapost.
ThisfieldisassignedtoALLposts,andcantakeuponeofthefollowingvalues:link,status,photo,video,offer[32].
Thismakescolumntextandimageanintersectionofprevioustwocolumns.Similarly,imageandnotextisa
http://precog.iiitd.edu.in/blog/2016/08/imagesononlinesocialmedia/
2/5
12/3/2016
APictureisWorth32.33Words:ImportanceofAnalyzingImagesonOnlineSocialMedia|PreCog'sBlog
subsetofcolumnwithimage,andtextandnoimageisasubsetofcolumnwithtext.Allvaluesinthetablein
parenthesisispercentagevalue.
Event
Total
Postswith
Postswith
Postswithtext
Postswithimage
Postswithtextand
posts
text
image
andimage
andnotext
noimage
22,820
6,868(30)
10,192(44)
538(2)
9,654(42)
6,330(28)
20,960
17,217(82)
7,463(36)
5,756(27)
1,707(8)
11,416(54)
67,453
28,030(42)
12,386(18)
1,553(2)
10,833(16)
26,477(39)
Eurocup2016
109,189
77,355(71)
61,119(56)
40,518(37)
20,601(19)
36,837(34)
Wimbeldon2015
111,417
80,469(72)
52,756(47)
37,862(34)
14,894(13)
42,607(38)
Parisattacks2015
131,548
78,803(60)
75,277(57)
32,861(25)
41,416(32)
45,942(35)
MalasiyanMH17
22,490
5,270(23)
2,947(13)
316(1)
2,631(12)
4,954(22)
IPL8cricket2015
48,329
31,526(65)
19,116(40)
9,251(19)
9,865(20)
22,275(46)
Gazaunrest2015
31,537
10,142(46)
6,157(20)
1,716(5)
4,441(14)
8,426(27)
AirAsiaflight
missing2014
Cricketworldcup
2015
Ebolaoutbreak
2014
crash2014
GiventhisgrowthofimagesandpicturesonOSM,andlessworkdoneontopicsrelatedtoOSM&images,thereisa
greatscopeforcontributinginthisdomain.TherearefullfledgedanddedicatedtraditionalconferenceslikeIEEE
InternationalConferenceonComputerVision,InternationalConferenceonMachineLearning(ICML),andIEEE
ConferenceonComputerVisionandPatternRecognition(CVPR)whichlookatimages.Thereneedssome
knowledgetransferfromtheseclassicdomainstoOSM.Itmayalsobethecasethat,inthepast,imageanalysiswas
notasadvancedasitisnow,so,advancementsinimageanalysis,includingneuralnetworksnowmakesitpossible
todosomereallycoolimageanalysiswhichcouldhavebeendifficultorimpossibletodoitearlier.Giventhelarge
amountofdataonOSM,andwithadvancedimageanalysistechniques,weshouldbeabletoanswersomevery
excitingresearchquestions.
SomespecifictopicsandproblemsthatIthinkthatwillbeinterestinginthisspaceofOSMandimages(thesearejust
myrandomthoughtsandtheyarenoncomprehensive):
Spreadofuntrustworthy/MisinformationonOSMthroughimages
Leakageofpersonalinformationlikecurrentlocation,etc.throughimagesonOSM
LeakageofsensitiveinformationlikeDOB,gender,etc.throughimagesonOSM
IfyouareinterestedinkeepingupdatedaboutouractivitiesatPrecog,youcanvisitourwebsiteorourFacbeook
pageIfyouhaveanysuggestionsorideastoexploreinthisdirection,feelfreetowritetome.
Acknowledgements:IthankmybrilliantstudentsPrateekDewan,NiharikaSachdeva,IndiraSen,KushagraSingh,
MeghaArora,HemankLamba,andVarunBharadhwajforhelpingwithputtingtogetherthesethoughts/some
numbers/analysisinthispost.ThankstoallmembersofPrecoggroupwheretheideaofstudyingimagesandtrying
itoutfromdifferentperspectivesstarted.
References
1.http://www.smartinsights.com/internetmarketingstatistics/happensonline60seconds/
2.http://www.businessinsider.com/werenowpostingastaggering18billionphotostosocialmediaeveryday
20145
3.http://www.adweek.com/socialtimes/twitterimagesstudy/493206
4.http://www.socialbakers.com/blog/1749photosmakeup93ofthemostengagingpostsonfacebook
5.https://people.csail.mit.edu/khosla/papers/www2014_khosla.pdf
http://precog.iiitd.edu.in/blog/2016/08/imagesononlinesocialmedia/
3/5
12/3/2016
APictureisWorth32.33Words:ImportanceofAnalyzingImagesonOnlineSocialMedia|PreCog'sBlog
6.PollyannaGonalves,MatheusArajo,FabrcioBenevenuto,andMeeyoungCha.2013.Comparingand
combiningsentimentanalysismethods.InProceedingsofthefirstACMconferenceonOnlinesocialnetworks
(COSN13).ACM,NewYork,NY,USA,2738.DOI=http://dx.doi.org/10.1145/2512938.2512951
7.TomerSimon,AvishayGoldberg,LimorAharonsonDaniel,DmitryLeykin,BruriaAdini.TwitterintheCrossFire
TheUseofSocialMediaintheWestgateMallTerrorAttackinKenya,PlosOne.
8.SarithaSK,DevshriroyD(2013)SemanticOrientationofSentimentAnalysisonSocialMedia.International
JournalofComputers&Technology11(4)24012409.
9.MunmunDeChoudhury,ScottCounts,andEricHorvitz.2013.PredictingPostpartumChangesinEmotionand
BehaviorviaSocialMedia.InProc.CHI13
10.MunmunDeChoudhury,ScottCounts,EricJHorvitz,andAaronHoff.2014.characterizingandpredicting
postpartumdepressionfromsharedfacebookdata.InProc.CSCW14.ACM,626638.
11.MunmunDeChoudhury,AndresMonroyHernandez,andGloriaMark.2014.NarcoEmotions:Affectand
DesensitizationinSocialMediaduringtheMexicanDrugWar.InProc.CHI14.ACM.
12.SatarupaGuha,TanmoyChakraborty,SamikDatta,MohitKumar,VasudevaVarma.TweetGrep:Weakly
SupervisedJointRetrievalandSentimentAnalysisofTopicalTweets.IntheproceedingsofICWSM2016.
13.SoroushVosoughi,DebRoy.ASemiAutomaticMethodforEfficientDetectionofStoriesonSocialMedia.Inthe
proceedingsofICWSM2016.
14.DavidAlvarezMelis,MartinSaveski.TopicModelinginTwitter:AggregatingTweetsbyConversations.Inthe
proceedingsofICWSM2016.
15.TimAlthoff,CristianDanescuNiculescuMizil,DanJurafsky.HowtoAskforaFavor:ACaseStudyonthe
SuccessofAltruisticRequests.IntheproceedingsofICWSM2014.
16.EfthymiosKouloumpis,TheresaWilson&JohannaMoore2011.TwitterSentimentAnalysis:TheGoodtheBad
andtheOMG!(ICWSM11)
17.AlexanderPakandPatrickParoubek2010.TwitterasaCorpusforSentimentAnalysisandOpinionMining.In
LREC,vol.10,pp.13201326.
18.AliakseiSeveryn,andAlessandroMoschitti.Twittersentimentanalysiswithdeepconvolutionalneural
networks.Proceedingsofthe38thInternationalACMSIGIRConferenceonResearchandDevelopmentin
InformationRetrieval.ACM,2015.
19.CceroNogueiradosSantos,andMairaGatti.DeepConvolutionalNeuralNetworksforSentimentAnalysisof
ShortTexts.COLING.2014.
20.DuyuTang,FuruWei,NanYang,MingZhou,TingLiu,andBingQin.LearningSentimentSpecificWord
EmbeddingforTwitterSentimentClassification.InACL(1),pp.15551565.2014.
21.1.Vaziripour,Elham,ChristopheGiraudCarrier,andDanielZappala.AnalyzingthePoliticalSentimentof
TweetsinFarsi.TenthInternationalAAAIConferenceonWebandSocialMedia.2016.
22.2.Peng,Nanyun,YimingWang,andMarkDredze.LearningPolylingualTopicModelsfromCodeSwitched
SocialMediaDocuments.ACL(2).2014.
23.3.Weerkamp,Wouter,SimonCarter,andManosTsagkias.Howpeopleusetwitterindifferentlanguages.
(2011):12.
24.4.Volkova,Svitlana,TheresaWilson,andDavidYarowsky.ExploringDemographicLanguageVariationsto
ImproveMultilingualSentimentAnalysisinSocialMedia.EMNLP.2013.
http://precog.iiitd.edu.in/blog/2016/08/imagesononlinesocialmedia/
4/5
12/3/2016
APictureisWorth32.33Words:ImportanceofAnalyzingImagesonOnlineSocialMedia|PreCog'sBlog
25.AnupamJamatia,BjornGambck,andAmitavaDas.2015.PartofSpeechTaggingforCodeMixedEnglish
HindiTwitterandFacebookChatMessages.ProceedingsofRecentAdvancesinNaturalLanguage
Processing,page239.
26.SujanKumarSaha,ParthaSarathiGhosh,SudeshnaSarkar,andPabitraMitra.2008.NamedEntity
RecognitioninHindiusingMaximumEntropyandTransliteration.ResearchjournalonComputerScienceand
ComputerEngineeringwithApplications,pp.3341.
27.AyushKumar,SarahKohail,AsifEkbal,andChrisBiemann.2015.IITTUDA:Systemforsentimentanalysisin
indianlanguagesusinglexicalacquisition.MiningIntelligenceandKnowledgeExploration,pages684693.
28.Gupta,A.,Lamba,H.,Kumaraguru,P.,andJoshi,A.FakingSandy:CharacterizingandIdentifyingFake
ImagesonTwitterduringHurricaneSandy.2ndInternationalWorkshoponPrivacyandSecurityinOnlineSocial
Media(PSOSM),inconjunctionwiththe22thInternationalWorldWideWebConference(WWW)(2013).
29.Gupta,A.,Lamba,H.,andKumaraguru,P.$1.00perRT#BostonMarathon#PrayForBoston:AnalyzingFake
ContentonTwitter.IEEEAPWGeCrimeResearchSummit(eCRS),2013.
30.https://www.facebook.com/americanmuslims1/photos/a.809524959106862.1073741828.527806667278694/990645217661501/?
type=1&theater
31.https://github.com/tesseractocr/tesseract
32.https://developers.facebook.com/docs/graphapi/reference/v2.7/post#read
33.https://www.facebook.com/ChristianChronicle/photos/a.83579936833.99565.11127431833/10153806013491834/?
type=3&theater
34.https://www.facebook.com/roberta.metsola/photos/a.406836966100205.1073741826.406824526101449/839065439544020/?
type=3&theater
35.https://www.facebook.com/516601545154233/photos/a.519535361527518.1073741828.516601545154233/574613042686416/?
type=3&theater
Like 37
Share
Tweet
Share
ThisentrywaspostedinResearchbyPK.Bookmarkthepermalink
[http://precog.iiitd.edu.in/blog/2016/08/imagesononlinesocialmedia/].
AboutPK
AssociateProfessor@IIITDelhihttp://precog.iiitd.edu.in
ViewallpostsbyPK
http://precog.iiitd.edu.in/blog/2016/08/imagesononlinesocialmedia/
5/5