Affective Computing and Sentiment Analysis: Emotion, Metaphor and Terminology

158 Pages·2011·2.218 MB·Text, Speech and Language Technology 45

by  Sam Glucksberg (auth.), Khurshid Ahmad (eds.)

AffectiveComputing andSentimentAnalysis
TEXT,SPEECHANDLANGUAGETECHNOLOGY
VOLUME45
SeriesEditor
NancyIde,VassarCollege,NewYork
EditorialBoard
HaraldBaayen,MaxPlanckInstituteforPsycholinguistics,TheNetherlands
KennethW.Church,MicrosoftResearchLabs,RedmondWA,USA
JudithKlavans,ColumbiaUniversity,NewYork,USA
DavidT.Barnard,UniversityofRegina,Canada
DanTufis,RomanianAcademyofSciences,Romania
JoaquimLlisterri,UniversitatAutonomadeBarcelona,Spain
StigJohansson,UniversityofOslo,Norway
JosephMariani,LIMSI-CNRS,France
Forfurthervolumes:
http://www.springer.com/series/6636
Affective Computing
and Sentiment Analysis
Emotion, Metaphor and Terminology
Editedby
Khurshid Ahmad
TrinityCollege,Ireland
123
Editor
KhurshidAhmad
TrinityCollege
SchoolofComputerScienceandStatistics
Dublin2
Ireland
[e-mail]
ISSN1386-291X
ISBN978-94-007-1756-5 e-ISBN978-94-007-1757-2
DOI10.1007/978-94-007-1757-2
SpringerDordrechtHeidelbergLondonNewYork
LibraryofCongressControlNumber:2011934267
(cid:2)c SpringerScience+BusinessMediaB.V.2011
Nopartofthisworkmaybereproduced,storedinaretrievalsystem,ortransmittedinanyformorby
anymeans,electronic,mechanical,photocopying,microfilming,recordingorotherwise,withoutwritten
permissionfromthePublisher,withtheexceptionofanymaterialsuppliedspecificallyforthepurpose
ofbeingenteredandexecutedonacomputersystem,forexclusiveusebythepurchaserofthework.
Printedonacid-freepaper
SpringerispartofSpringerScience+BusinessMedia(www.springer.com)
Acknowledgments
I must begin by acknowledging the patience and goodwill of the contributors of
this book. Many of the authors traveled to Marrakesh to participate in the Work-
shop on Emotion, Metaphor, Ontology, and Terminology (EMOT) in August 2008
from places as far as the USA, Hong Kong and the EU. This Workshop was part
of the bi-annual, EU-sponsored Language Resources and Evaluation Conference
(LREC).SelectedpapersfromtheWorkshopwererevisedbythecontributorsduring
2009–2010.EquallypatientandhelpfulweremyeditorsJolandaVoogdandHelen
van der Stelt of Springer Verlag. I am grateful to all the authors and editors. Last
butnotleast,mygratefulthankstomyProjectManager,SangeethaSathiamurthyat
Springer’s,forallthehelpshehasprovidedduringthefinalstagesoftheproduction
ofthisbook.
Papers presented in this book were reviewed by the following colleagues:
Gerhard Budin (Universität Wien, Vienna), Ann Devitt (Trinity College, Dublin),
SamGlucksberg(PrincetonUniversity),GerdHeyer(UniversitätLeipzig),Mícheál
Mac An Airchinnigh (Trinity College, Dublin) Maria Teresa Musacchio (Univer-
sityofPadova),MariaTeresaPazienza(UniversityofRomaTorVergata),Margaret
Rogers(UniversityofSurrey,UK),CarlVogel(TrinityCollege,Dublin),TonyVeal
(UniversityCollege,Dublin),andYorickWilks(UniversityofSheffield,UK).The
whole collection was reviewed by anonymous referees selected by my editors
JolandaandHelen.Mygratefulthankstoallthenamedandanonymousreferees.
My colleagues, Andrea Zemánková, and Chaoxin Zheng at Trinity College,
helped in formatting the initial draft of this book. Thanks are due to my doctoral
students, Aaron Gerow and Daniel Isemann, who formatted the final draft of this
collectionandhelpedinproofreading.
The support of Long Room Hub Project, Trinity College, Dublin, Ireland, is
gratefullyacknowledged.
v
Introduction: Affect Computing and Sentiment
Analysis
KhurshidAhmad
Many of the things we think about, actions we take, the way we react to stimuli,
generateafeelingorsubjectiveexperience,forexample,anemotion,oramood.The
generictermusedinthetwentiethcenturypsychologyandphilosophyliteratureto
denotesuchanemotionormoodisanold,MiddleEnglish(fourteenthcentury)word
affect. The outward display of affect is manifested typically by facial expressions
and postures by humans. Language plays a key role in the articulation of affect
bothinspokenlanguageandinwrittentext:moreinterestinglyperhapsistheuseof
metaphorsinthearticulationofemotion,moodandsentiment.Theuseofmetaphors
toexpressaffectisnotrestrictedtotheimaginativegenreoffictionandcanbefound
inspecialistlanguagesofscienceingeneral[11],andinnatural[12]andbiological
sciences[17]inparticular.Financeandeconomicshavetheirfairshareofmetaphors
forexpressingaffect.
Much in the vein of doctrine of affects in musical aesthetics, that music can be
composed to arouse a variety of specific and involuntary emotions in the listener,
behaviouralpsychologistshavearguedthatanoutwarddisplayofspecificemotions
canbeprecipitatedbyframingpoliticalorfinancialpropositionsinaparticularway
[8].Themetaphoricalmeltdownoftheglobalfinancialsystemsin2008wasprecip-
itated,inpart,bystrong,negativesentimentaboutfinancialsystemsinnewsreports
andeditorialspennedbyjournalists,inop-edcolumnswrittenbycommentatorsand
experts,inblogsbycitizens,andinspeechesbybankersandbankregulatorsalike.
Equally important in the meltdown was the contribution of the actively promoted,
positive sentiment in reports and comments on the so-called riskless investments;
these texts comprised hyperbole like collateral obligations and inelegant coinages
likeliarsloans.
The2008financialcrisis,andmanybeforethat,showshowfigurativelanguage
can be used to engender a false sense of security prior to the crisis and then con-
structsinthesamelanguagecanbeusedtopacifythevictimsofthecrisisafterward.
The use of figurative language to describe the changes in financial markets has a
long pedigree: In the seventeenth century we had financial mania, the eighteenth
K.Ahmad(B)
SchoolofComputerScienceandStatistics,TrinityCollege,Dublin2,Ireland
e-mail:[e-mail]
vii
viii Introduction:AffectComputingandSentimentAnalysis
hadfinancialbubbles,thenineteenthhadfinancialpaniconlytobereplacedbythe
emotivetermfinancialmarketcrashandthepsycho-babblisheconomicdepression
in the first half of twentieth century [4]. The dominance of physical metaphors in
the second half of the twentieth century begat the wave-like economic recession.
Climaticmetaphorswereusedinthetwenty-firstcentury,forexamplecreditfreeze,
onlytobereplacedbythestronger-closer-to-the-factstermslikecreditsqueezeand
creditcrunch.Thesetermsareadaptedbyscholarsinfinanceandotherdisciplines,
so,forexample,financialpanichasledtothetermbankingpanicinmodernfinance
literature[3,13].Thecreativityofexpertsinfinanceandtheirsiblingsonandaround
thetradingfloorscontinuesapace:therearehaircutstodescribeinvestorlosses[15]
andtheintriguingdarkpoolsdenotinganewtypeofexchangedesignedtoaddress
theproblemsthatarisefromthetransparentnatureofatypicalstockexchange[9]!
Therollingglobalnews,usuallyavailabledigitally,andtheeasewithwhichopin-
ions can be expressed about such news, is not limited to economics or finance, to
politicsorscience,butnowencompasseseverywalkoflife.Thisnew‘democracy’,
which allows almost anybody with access to a digital device (computers, mobile
phones) to express their opinions, means a proliferation of vast amounts of digital
text and speech. This large volume of news and views, often comprising emotion
and mood, cannot be analysed for its subjective contents by humans within a rea-
sonabletimeifatall.Thesearchfor,andtheimportanceof,feelingsandsubjective
experience articulated in language, has led to the coinage of the term sentiment
analysis:asystematic,computer-basedanalysisofwrittentextorspeechexcerpts,
for extracting the attitude of the author or speaker about a specific topic. Indeed,
the origin of the sentiment analysis lies in the mid-twentieth century political sci-
ence, where pioneers like Harold Lasswell [10] and Philip Stone [14], developed
content-analysissystemsforanalysingspeechesbypoliticiansandthemanifestoof
political parties. Stone created the General Inquirer system, and his dictionary of
affectwords,with82differentcategoriesofaffect,isstillusedtoday[1,16].
Thegoalofthecomputerscientistsinvolvedinbuildingsuchsentimentanalysis
systems,istoidentifyandtabulatethejudgementoftheauthor/speakerortodiscern
howmuchorhowlittletheauthorspeakervaluesanobject,event,state,orabstract
idea. The longer term aim of such an analysis is to identify and tabulate what, if
any,involuntaryemotionalreactionareader/listenermayhavetoanauthor/speaker.
Indeed, sentiment analysis aka opinion mining, is used extensively in brand man-
agement and product promotion [2]. In economics and finance, sentiment analysis
enthusiasts suggest that in addition to fundamental analysis (the analysis of the
assets of an enterprise) and technical analysis (the analysis of the changes in the
valuesofshareorcommodityprices),itisessentialtousesentimentanalysis–the
inclusion of rumours in the market place related to an impending announcement
of a break through or meltdown, in order to explain phenomenon that cannot be
explained by fundamental and technical analysis. These phenomena include the
technology induced booms and busts, like the dot com boom/bust, and spikes in
commodityprices.
Language, as it appears to the contributors in this volume, is the key to senti-
ment analysis. Affect is articulated in text and speech through an exciting use of
Introduction:AffectComputingandSentimentAnalysis ix
metaphors.SamGlucksberg[5]hasusedthemetaphors–myspouse’slawyerisa
sharkandmyjobismyjail–toexplainhowlanguageisusedtorepresentsomething
asrepresentativeorsuggestiveofsomethingelse.ForAndrewGoatly,theubiquity
ofmetaphorineverydaylanguageandinspeciallanguagesofscience,technology,
andthearts,suggeststhat’farfrombeingananomaly’,metaphors’become’basic
[6].Cognitivescientists,likeJerryHobbs,motivatedby’thecomputermetaphorof
the mind’ [7], have attempted to address the notions of quantitative semantics by,
forexample,lookingatspatialrelationships.YorickWilks,apioneerofword-sense
disambiguation and information extraction [18], links the enterprise of sentiment
analysis to be a synthesis of work carried out in artificial intelligence, natural lan-
guageprocessing,computationallinguisticsandcontentanalysis(thisvolume).
In this collection of papers the contributions fall into three interacting cate-
gories–creationofmetaphors,affecttransferinconversation,andsentimentanal-
ysis systems in economics, finance and marketing. The epilogue of this collection
envisions futureinformationextraction systemsthatmayextractaffectinfreetext
andspeech:
I. Language use and the creation of metaphors. Here we have Sam Glucks-
berg’scontributiononthecreationofnewcategoriesusingmetaphors,Andrew
Goatly’sonhowmetaphorsareusedinconceptualisationandintheexpression
ofopinions,andJerryHobbsandAndrewGordon’sexplorationofthesemantic
basisofmetaphors.CarlVogelsynthesisesthenotionsofmetaphor formation
and belief revision for arguing that metaphors and generics are species of the
samekind–hispaperalsoprovidesalinktotheanalysisofemotivelanguage
use.
II. Affecttransferinconversationanduncertainty.AlanWalington,JohnBarnden
and colleagues describe an innovative approach to affect transfer in conversa-
tional situations and Rieks op den Akker and colleagues deal with detecting
‘uncertaintyinspokendialogues’.
III. Sentimentanalysisineconomics,financeandmarketing.
EconomicsandFinance:MariaTeresaMusacchiolooksatmetaphorsused
in economics in English and Italian and asks whether metaphors are
universal or culture specific. Moshe Koppel and colleagues describe
how supervised learning methods can be used to learn the association
between polarity of financial news and key financial indicators. Khur-
shidAhamdintroducestheeconometricnotionsofreturnandvolatility
for measuring changes in sentiment as articulated in (financial) news:
these changes insentiment arethen correlated withimportantfinancial
indicatorsforanalysisandprediction.
MarketingandProductReviews:GerdHeyerandassociateslookatsenti-
mentanalysisinmarketing.MarcBoullétogetherwithDamienPoirier
and colleagues, compares and contrasts the relative merits of machine
learningandlinguisticanalysistechniquesinpredictingtheopinionsof
reviewersonmovieablogsite.
x Introduction:AffectComputingandSentimentAnalysis
The epilogue of this collection of papers on has been written by Yorick Wilks.
He argues that the interaction of contents analysis, artificial intelligence, particu-
larly natural language processing, cognitive sciences, and information extraction
mayhelpusinunderstandinghowmetaphoriscommunicatedbyhumansandmay
helpusinbuildingmachinesthatcanprocessmetaphorsandpossiblysentiments.
References
1. Ahmad, Khurshid. 2008. Edderkoppspinn eller nettverk: News media and the use of polar
wordsinemotivecontexts.Synaps21:20–36.
2. Candillier,Laurent,FrankMeyer,andM.Boullé.2007.Comparingstate-of-the-artcollabora-
tivefilteringsystems.InternationalconferenceonMachineLearningandDataMiningMLDM
2007,Leipzig/Germany.LectureNotesinComputerScience,2007,Volume4571/2007:548–
562.Berlin:Springer.
3. Chari,V.V.,and RaviJagannathan. 1988. Bankingpanics,information, andrationalexpec-
tationsequilibriumbankingpanics,information,andrationalexpectationsequilibrium.The
JournalofFinance43(3):749-761,(July,1988).
4. Galbraith, John K. 1958/1998. The affluent society – The fortieth anniversary edition.
NewYork:HoughtonMifflinCompany.
5. Glucksberg, Sam. 2001. Understanding figurative language: From metaphors to idioms.
Oxford:OxfordUniversityPress.
6. Goatly,Andrew.1997.Thelanguageofmetaphors.London:Routledge.
7. Hobbs,JerryR.1990.Literatureandcognition.Lecturenotes,number21,CenterfortheStudy
ofLanguageandInformation,Stanford,California.
8. Kahneman, Daniel, and Tversky, Amos. 2000. Choices, values, and frames. In Choices,
Values,andFrames,eds.D.KahnemanandA.Tversky,1–16.Cambridge,NY:Cambridge
UniversityPress.
9. Kuzman,G.,Nevmyvaka,Y.,Kearns,M.,and,J.W.Vaughan.2010.Censoredexplorationand
thedarkpoolproblem.CommunicationsoftheACM.53(5):99–107,(May2010).
10. Lasswell,HaroldD.,NathanLeitesandAssociates,eds.1949.Languageofpolitics:Studies
inquantitativesemantics.NewYork:GeorgeW.Stewart,Publisher,Inc.(Republishedin1965
bytheMITPress,Cambridge,MA).
11. Miller,A.L.1984.Imageryinscientificthought.Boston,Basel,Stuttgart:Birkhaeuser.
12. Pullman,B.1988.Theatominthehistoryofhumanthought.Oxford:OxfordUniversityPress.
13. Richardson,G.,andW.Troost.2009.Monetaryinterventionmitigatedbankingpanicsdur-
ingthegreatdepression:Quasi-experimentalevidencefromafederalreservedistrictborder,
1929–1933.JournalofPoliticalEconomy117(6):1031–1073,(December2009).
14. Stone,PhilipJ.,DexterC.Dunphy,MarshallS.Smith,DanielM.Ogilvie,andAssociates.
1966. The general inquirer: A computer approach to content analysis. Cambridge, MA:
TheMITPress.
15. Sturzenegger,F.,andJ.Zettelmeyer.2008.Haircuts:Estimatinginvestorlossesinsovereign
debtrestructurings,1998–2005.JournalofInternationalMoneyandFinance27(5):780–805,
(September2008).
16. Tetlock,PaulC.2007.Givingcontenttoinvestorsentiment:Theroleofmediainthestock
market.JournalofFinance62:1139–1168.
17. Verschuuren,G.M.1986.Investigatingthelifesciences:Anintroductiontothephilosophyof
science.Oxford:PergamonPress.
18. Wilks,Y.2008.Onwhoseshoulders?ComputationalLinguistics34(4):1–16.