3/9/2017 HowtoWriteanAbstract

HowtoWriteanAbstract
PhilipKoopman,CarnegieMellonUniversity
October,1997

Abstract
Becauseonlinesearchdatabasestypicallycontainonlyabstracts,itisvitaltowriteacompletebutconcise
descriptionofyourworktoenticepotentialreadersintoobtainingacopyofthefullpaper.Thisarticledescribes
howtowriteagoodcomputerarchitectureabstractforbothconferenceandjournalpapers.Writersshould
followachecklistconsistingof:motivation,problemstatement,approach,results,andconclusions.Following
thischecklistshouldincreasethechanceofpeopletakingthetimetoobtainandreadyourcompletepaper.

Introduction
Nowthattheuseofonlinepublicationdatabasesisprevalent,writingareallygoodabstracthasbecomeeven
moreimportantthanitwasadecadeago.Abstractshavealwaysservedthefunctionof"selling"yourwork.But
now,insteadofmerelyconvincingthereadertokeepreadingtherestoftheattachedpaper,anabstractmust
convincethereadertoleavethecomfortofanofficeandgohuntdownacopyofthearticlefromalibrary(or
worse,obtainoneafteralongwaitthroughinterlibraryloan).Inabusinesscontext,an"executivesummary"is
oftentheonlypieceofareportreadbythepeoplewhomatteranditshouldbesimilarincontentifnottonetoa
journalpaperabstract.

Checklist:PartsofanAbstract
Despitethefactthatanabstractisquitebrief,itmustdoalmostasmuchworkasthemultipagepaperthat
followsit.Inacomputerarchitecturepaper,thismeansthatitshouldinmostcasesincludethefollowing
sections.Eachsectionistypicallyasinglesentence,althoughthereisroomforcreativity.Inparticular,theparts
maybemergedorspreadamongasetofsentences.Usethefollowingasachecklistforyournextabstract:

Motivation:
Whydowecareabouttheproblemandtheresults?Iftheproblemisn'tobviously"interesting"itmightbe
bettertoputmotivationfirstbutifyourworkisincrementalprogressonaproblemthatiswidely
recognizedasimportant,thenitisprobablybettertoputtheproblemstatementfirsttoindicatewhich
pieceofthelargerproblemyouarebreakingofftoworkon.Thissectionshouldincludetheimportanceof
yourwork,thedifficultyofthearea,andtheimpactitmighthaveifsuccessful.
Problemstatement:
Whatproblemareyoutryingtosolve?Whatisthescopeofyourwork(ageneralizedapproach,orfora
specificsituation)?Becarefulnottousetoomuchjargon.Insomecasesitisappropriatetoputthe
problemstatementbeforethemotivation,butusuallythisonlyworksifmostreadersalreadyunderstand
whytheproblemisimportant.
Approach:
Howdidyougoaboutsolvingormakingprogressontheproblem?Didyouusesimulation,analytic
models,prototypeconstruction,oranalysisoffielddataforanactualproduct?Whatwastheextentof
yourwork(didyoulookatoneapplicationprogramorahundredprogramsintwentydifferent
programminglanguages?)Whatimportantvariablesdidyoucontrol,ignore,ormeasure?
Results:
What'stheanswer?Specifically,mostgoodcomputerarchitecturepapersconcludethatsomethingisso
manypercentfaster,cheaper,smaller,orotherwisebetterthansomethingelse.Puttheresultthere,in

https://users.ece.cmu.edu/~koopman/essays/abstract.html 1/3
3/9/2017 HowtoWriteanAbstract

numbers.Avoidvague,handwavingresultssuchas"very","small",or"significant."Ifyoumustbe
vague,youareonlygivenlicensetodosowhenyoucantalkaboutordersofmagnitudeimprovement.
Thereisatensionhereinthatyoushouldnotprovidenumbersthatcanbeeasilymisinterpreted,buton
theotherhandyoudon'thaveroomforallthecaveats.
Conclusions:
Whataretheimplicationsofyouranswer?Isitgoingtochangetheworld(unlikely),beasignificant
"win",beanicehack,orsimplyserveasaroadsignindicatingthatthispathisawasteoftime(allofthe
previousresultsareuseful).Areyourresultsgeneral,potentiallygeneralizable,orspecifictoaparticular
case?

OtherConsiderations
Anabstractmustbeafullyselfcontained,capsuledescriptionofthepaper.Itcan'tassume(orattemptto
provoke)thereaderintoflippingthroughlookingforanexplanationofwhatismeantbysomevaguestatement.
Itmustmakesenseallbyitself.Somepointstoconsiderinclude:

Meetthewordcountlimitation.Ifyourabstractrunstoolong,eitheritwillberejectedorsomeonewill
takeachainsawtoittogetitdowntosize.Yourpurposeswillbebetterservedbydoingthedifficulttask
ofcuttingyourself,ratherthanleavingittosomeoneelsewhomightbemoreinterestedinmeetingsize
restrictionsthaninrepresentingyoureffortsinthebestpossiblemanner.Anabstractwordlimitof150to
200wordsiscommon.
Anymajorrestrictionsorlimitationsontheresultsshouldbestated,ifonlybyusing"weaselwords"such
as"might","could","may",and"seem".
Thinkofahalfdozensearchphrasesandkeywordsthatpeoplelookingforyourworkmightuse.Besure
thatthoseexactphrasesappearinyourabstract,sothattheywillturnupatthetopofasearchresult
listing.
Usuallythecontextofapaperissetbythepublicationitappearsin(forexample,IEEEComputer
magazine'sarticlesaregenerallyaboutcomputertechnology).But,ifyourpaperappearsinasomewhat
untraditionalvenue,besuretoincludeintheproblemstatementthedomainortopicareathatitisreally
applicableto.
Somepublicationsrequest"keywords".Thesehavetwopurposes.Theyareusedtofacilitatekeyword
indexsearches,whicharegreatlyreducedinimportancenowthatonlineabstracttextsearchingis
commonlyused.However,theyarealsousedtoassignpaperstoreviewcommitteesoreditors,whichcan
beextremelyimportanttoyourfate.Somakesurethatthekeywordsyoupickmakeassigningyourpaper
toareviewcategoryobvious(forexample,ifthereisalistofconferencetopics,useyourchosentopic
areaasoneofthekeywordtuples).

Conclusion
Writinganefficientabstractishardwork,butwillrepayyouwithincreasedimpactontheworldbyenticing
peopletoreadyourpublications.Makesurethatallthecomponentsofagoodabstractareincludedinthenext
oneyouwrite.

FurtherReading
Michaelson,Herbert,HowtoWrite&PublishEngineeringPapersandReports,OryxPress,1990.Chapter6
discussesabstracts.

Cremmins,Edward,TheArtofAbstracting2ndEdition,InfoResourcesPress,April1996.Thisisanentirebook
aboutabstracting,writtenprimarilyforprofessionalabstractors.

https://users.ece.cmu.edu/~koopman/essays/abstract.html 2/3
3/9/2017 HowtoWriteanAbstract

Copyright1997,PhilipKoopman,CarnegieMellonUniversity.
Embeddedsystemdesignersmaybeinterestedinmyblog.

https://users.ece.cmu.edu/~koopman/essays/abstract.html 3/3