Relative Trends in Scientific Terms on Twitter

Rela%veTrendsinScien%ficTermsonTwi4er

VictoriaUren,Aba‐SahDadzieTheOAKGroup,Dept.ofComputerScience,TheUniversityofSheffield

altmetrics11:TrackingscholarlyimpactonthesocialWeb

Introduc%on

•  scien%ficresearchtradi%onallydisseminatedviajournals,books,scien%ficconferences

•  newformofdiscourse–onlinesocialmedia–  suitableforumfordissemina%ngscien%ficresearch?

–  doscien%stsengagewithonlinesocialmedia?

–  aretheresufficientamountsofinforma%ononscien%fictopics?

•  aretheresuitablemetricsformeasuringscien%ficimpactonline?–  betweenscien%sts?–  forpublicengagement?

•  arethesenewmeasurescomparabletoformalmetrics?


Outline

•  Aims/Introduc%on•  RelatedWork

•  Experiment–  Data–  Analysis&Results

•  Conclusions

•  NextSteps

•  Acknowledgements


Outline

•  Aims/Introduc%on

•  RelatedWork


•  Conclusions

•  NextSteps



RelatedWork

•  Garfield,E.(from1950s)–  fatherofscientometrics

•  Priemetal.(2010)–  Scientometrics2.0asanewmetricformeasuringscholarlyimpactonsocialweb

•  Lane(2010)–  needtoimprovemetricsusedtomeasurescien%ficimpact

•  Micheletal.(2011)–  GooglenGramstoanalyseculture–  a.o.,recognisedfameforscien%stslow…

•  Cheongetal.(2009)–  H1N1spike(trend)detectedonTwi4erduringflupandemic(May2009)

•  Roweetal.(2011)–  influenceofcontentandauthorfeaturesonpredic%onofac%ve,longterm

discussionsonsocialweb•  Kinsellaetal.(2011)

–  usinghyperlinkedmetadatatoaidcategorisa%onoftopicsdiscussedinonlinesocialmedia


Outline



•  Conclusions

•  NextSteps



Experiment

•  exploratoryexperiment–  todeterminefrequencyofoccurrenceofscien%fictermusagein

onlinesocialmedia

•  dataset–  threesetsof(scien%fic)termsselectedfromUNESCOthesaurus–  GoogleBooksNGramscorpususedasabaseline

–  300tweetscollectedineachsample,usingTwi4erAPI,forselectedterms

•  frequency/usageanalysis


Outline


•  Experiment

– Data–  Analysis&Results

•  Conclusions

•  NextSteps



UNESCOThesaurus1GramTerms

Topic TermsPhysicalSciences Ioniza%on,Electromagne%sm,Crystallography

ChemicalSciences Phosphorus,Alkalinity,Microchemistry

EarthSciences Permafrost,Lithosphere,Glaciology

•  selec%oncriteria–  minimisa%onofnoiseduetopolysemy

–  avoidanceofscien%fictermswithothercommon/colloquialusage

–  termsuniquetoapar%culartopic

–  wordswithasinglestem

–  1Gramsonly


BaselineDataset–Google1Grams

•  obtainedfromGoogleBooksNGramscorpus1

•  totalNGramsbyyearforthreesetsofterms–  2006–116,029–  2007–126,206–  2008–111,417

•  annualvaria%onbytopic(oftotalNGramsbaselinedataset)–  ChemicalSciences50‐60%

–  PhysicalSciences30‐40%–  EarthSciences~10%

•  [1]h4p://ngrams.googlelabs.com/datasets


BaselineDataset–Google1Grams


Twi4erDataset

SampleID CollecAonPeriod ElapsedTime(h)

T‐300‐1 TueMar0120:56:43GMT2011–ThuMar0314:22:18GMT2011

41

T‐300‐2 FriMar0402:35:55GMT2011–SunMar0618:38:05GMT2011

64

T‐300‐3 MonMar0720:31:11GMT2011–WedMar0916:21:36GMT2011

44

•  threesamplescollected,containing300consecu%vetweetseach•  ~0.003%oftotaltweetsovercollec%onperiod


Outline


•  Experiment–  Data

– Analysis&Results

•  Conclusions

•  NextSteps



Twi4erc.f.GoogleNGrams



•  highervaria%onindistribu%onforTwi4ersample–  howeverlargelyinlinewithGoogleNGrams

•  canGoogleNGramsserveasasuitablebaseline?–  needtomorecloselyexaminevaria%on…

•  notablepeaksinTwi4ersampleforthreeterms–  Permafrost(EarthSciences)

–  Alkalinity(ChemicalSciences)–  Phosphorus(ChemicalSciences)

•  arethesepoten%altrends?




Twi4erc.f.GoogleNGrams•  Permafrost

–  17%and15%inTwi4ersamples(T‐300‐1&2)–c.f.5%inG‐2006‐2008–  41outof113tweets(36%)usedinscien%ficcontext–  largenumberoftweetsreferredto

•  onlinegameserver1•  designercaseforiPhone

•  Alkalinity–  nonefoundtohavescien%ficcontent–  mostlyusedinpseudo‐scien%fichealthadvice–  peakinT‐300‐2(31outof60tweets–~50%)

•  dominatedbypHmeasuresinswimmingpools&fishtanks•  influenceprobablyduetocollec%onperiod–weekend–engagementinleisureac%vi%es

•  [1]h4p://www.everquest2.com/Permafrost


ExampleTweets–Permafrost

•  advert/chat–  @HDNinjacpgotoPermafrostItsneverfull:FriMar0405:21:02GMT2011–  @Riffy8888heyCouldyouCometomyPartybirthdayPartyonCPMarch13Server

PermafrostDock6:00PST:SunMar0604:37:13GMT2011–  PartyServerPermafrostDockPleaseGoIt'sAnEarlyBirthdayPartyForme:ThuMar03

01:38:28GMT2011

•  cold–  36inchesofpermafrosts%ll,Iwanttostakemybirdcondob4thesquirralsknockit

downagain..bas%ds..allof'm:SatMar0501:51:48GMT2011

•  science–  FireandIce:PermafrostMeltSpewsCombus%bleMethaneh4p://%ny.ly/be8q:FriMar

0416:43:10GMT2011–  (retweeted)‐ExpertsMonitorMethaneReleasefromPermafrost:Overthepastfew

years,methanelevelsaroundtheworldhaveb...h4p://bit.ly/hvVEJX:WedMar0212:27:25GMT2011

–  RT@NetNewsBuzz:PermafrostMeltSoonIrreversibleWithoutMajorFossilFuelCutsh4p://%nyurl.com/5w8w2oh#oil#climate#CO2#fossilfuels:ThuMar0302:57:48GMT2011


ExampleTweets–AlkalinityT‐300‐2

•  ChemistryHelpNeeded!pH,concentra%onofcarbonatespeciesandalkalinity...justgotpublished:h4p://bit.ly/hUCpz7–  URLpointstotheques%onon“MyChemistryTutor”–homework?

•  retweeted–  Thepropertotalalkalinityforyourpoolis100ppm.h4p://su.pr/8hrxCE:FriMar

0419:02:20GMT2011–  IftheTotalAlkalinityinyourswimmingpoolislow,yourpHwillbelow.h4p://

su.pr/8hrxCE:FriMar0420:34:11GMT2011

•  spam/adverts(includingretweets)–  @Poet_Carl_Wa4s:somefoodscreateacidityoralkalinityayerthey‚Äôre

metabolized...h4p://ping.fm/GQTvA#KnowledgeIsPower!:SatMar0502:38:55GMT2011

–  RT@CourtneyPool:Greenjuice,ohLiquidEmeraldElixirofLifeandAlkalinity!CoursethroughmyBODY!#juicing:SunMar0618:34:29GMT2011


SampleID

Total LegislaAon NutriAon OtherSciences

Industry WhitePhosphorus

T‐300‐1 129 46 16 29 4 5

T‐300‐2 119 4 26 35 9 5

T‐300‐3 171 12 23 37 42 19

•  Twi4ertrendsforPhosphorusinsampleT‐300‐3–  Industry

•  takeoverofaBraziliancompanybytheIndianfirmUnitedPhosphorus–  WhitePhosphorus

•  17retweetsofanemo%vemessage(rela%ontoMiddleEastwars)

Twi4erc.f.GoogleNGrams:Phosphorus


Twi4erc.f.GoogleNGrams:Phosphorus

•  usagelargelywithscien%ficcontent–  withrela%onships,a.o.,tolegal,nutri%onal&economiccontext–  fivemaincategoriesiden%fied

•  Legisla%on–  limitstouseinfer%liser,soap

•  Nutri%on–  phosphoruscontent

•  OtherScience–  peakphosphorus,pollu%on–  discoveryofarsenicreplacingphosphorusinamicrobe–  tweetsaboutnewpaperonRedfieldra%oinorganisms

•  Industry–  mergers,pricesofPhosphorus‐containinggoods

•  WhitePhosphorus–  useinMiddleEastwars


ExampleTweets–Phosphorus •  Legisla%on

–  RT@YarnPlayCafe:ThefactthathewantstorepealthephosphorusbanandkilltheMadisonlakesis,byitself,enoughto#killthisbill...:TueMar0802:04:49GMT2011

•  Nutri%on–  Big,wetsnowflakesdriyoverthefarm.Towarmup,ItrysomeHorlicks,awheat/barley/whey

drinkwithlotsofcalcium&phosphorus.Mmmm.:TueMar0120:56:43GMT2011–  VitaminDactsasanhormoneandplaysacontrollingroleinthemetabolismofcalciumand

phosphorus:SunMar0612:12:36GMT2011•  OtherScience

–  [java]129:GreaterPhosphorusEfficiencyh4p://bit.ly/iehsmK#agriculture:WedMar0214:36:21GMT2011

•  Industry–  #stocks#bse#nseBuyUnitedPhosphorus‐posi%vemovetotaplargestLa%nAmericanmarket;

Edelweissh4p://dlvr.it/JdSpV:TueMar0817:22:55GMT2011–  Enshi:Wugangdevelopstechniquetohandlehigh‐phosphorusironore‐SteelBusinessBriefing

(subscrih4p://uxp.in/30538045:TueMar0809:33:05GMT2011•  WhitePhosphorus

–  DearAmerica,yourwhitephosphorusanddepleteduraniumcannotstopthegrowthofIraq'sfuture.IraqWillRise.:WedMar0207:49:21GMT2011

–  @Remroumsofirsttheystealourland,nowtheywantour"tac%cs"i.e.poetry?iguessthewhitephosphorusjustisn'tcu�ngitanymore.:SatMar0503:42:44GMT2011

•  ???–  @p_kojo‐PhosphorusPotassium‐Pinocchio,I'msogladwefoundeachothernwwecanhav

lotsoffun:):SunMar0613:43:10GMT2011


Outline



•  Conclusions•  NextSteps



Conclusions–Experiment•  recognisedchallenges

–  baselinecorpusforonlinesocialmediadifficulttoobtain•  verysmall(rela%vely)samplesfoundinTwi4erstream

•  difficulttoobtainrepresenta%vesamples moreeffec%vemethodsrequiredtoextractlowerfrequencyterms

–  difficultyreproducingexperiments

–  reliability,ethical&privacyissues–duetouser‐createdcontent

•  whatisasuitable,publiclyavailablebaselinecorpus?–  GoogleNGrams?

•  differentinforma%oncollec%onmethodsfromonlinesocialmedia–  coverageoftopicsmayseelargevaria%onbetweencorpora

–  anyothers?•  Wikipedia/DBpedia?TREC?


EngagementwiththeWeb?

•  whydoscien%stsnottweet?(orengagemuchinothersocialmedia)?–  isthewebnotseentoenforcesufficientscien%ficrigour?

–  doscien%stsnotviewthewebasapoten%alaudience?•  isthewebaudienceasuitablepeerreviewer?

•  whydoscien%stshesitatetodisseminateinforma%ononline?–  poten%alforideastobestolen?–  trust–howtodifferen%atebetweenvalidscienceandpseudo‐science,

spamandadverts?

•  socialmedialargelydrivenbypersonalinterest,sen%ment,opinion–  mayexplainlowscien%ficcontent

–  morecolloquialuseofwhatistradi%onallyscien%ficterminology


Implica%onsforAltmetrics•  however‐somelevelofscien%ficdiscourseonTwi4er

–  e.g.,Phosphorusiden%fiedasapoten%alTwi4ertrend

•  onlinesocialmediamays%llhavepoten%altoserveasanaltmetricformeasuringimpactofscience

•  star%ngfromscientometrics‐whichlooksatauthorfeatures,e.g.,–  co‐cita%on–  affilia%on–rela%onshiptoreputa%on

•  correspondingfeaturesinonlinesocialmedia–  followers–  retweets–rela%onshiptotrust?


Outline



•  Conclusions

•  NextSteps



NextSteps•  replicateexperimentswithlargersamplesoverlongerperiod

–  moredetailedanalysis•  e.g.,hashtaganalysis;urlswithintweets•  focusontermswithmoretrendingpoten%al,e.g.,nanostructures,nanosilver

•  considerspecifictweets–  fromscien%ficmediaandjournals–  postedduringscien%ficconferences,congresses

•  comparisonwithotherindependentbaselinedatasets

•  compareTwi4erusewithindifferentdisciplines–  influenceofinterdisciplinarycollabora%ononuseofonlinesocialmedia?

•  createnewbenchmarksdata&experiments  definealt‐metricforscien%fictermusageinonlinesocialmedia


Acknowledgements

•  ElizabethCanofordiscussionsoncollec%onanduseofdatafromTwi4erstreams

•  V.S.Uren&A.‐S.Dadziefundedby:–  EuropeanCommission7thFrameworkProgrammeproject

SmartProducts(grantno.231204)


References

•  Garfieldbib‐h4p://garfield.library.upenn.edu/pub.html

•  Ma4hewRowe,SofiaAngeletouandHarithAlani.(2011)Predic%ngDiscussionsontheSocialSeman%cWeb,Proc.,ESWC(2)2011:405‐420

•  SheilaKinsella,MengjiaoWang,JohnBreslinandConorHayes.(2011)ImprovingCategorisa%oninSocialMediausingHyperlinkstoStructuredDataSources,Proc.,ESWC(2)2011:390–404

•  othersinpaperreferences–seeh4p://altmetrics.org/altmetrics11/uren‐v0

Relative Trends in Scientific Terms on Twitter

Documents