Proceedings of the Second ACL Workshop on Statistical Machine Translation

281 Pages·2007·6.68 MB·English

ACL 2007
PRAGUE
A C L 2 0 0 7
Proceedings of the Second Workshop
on Statistical Machine Translation
June 23, 2007
Prague, Czech Republic
The Association for Computational Linguistics
ProductionandManufacturingby
Omnipress
2600AndersonStreet
Madison,WI53704
USA
(cid:13)c 2007AssociationforComputationalLinguistics
OrdercopiesofthisandotherACLproceedingsfrom:
AssociationforComputationalLinguistics(ACL)
209N.EighthStreet
Stroudsburg,PA18360
USA
Tel: [phone]
Fax: [phone]
[email protected]
ii
Preface
The ACL 2007 Workshop on Statistical Machine Translation (WMT-07) took place on Saturday, June
23 in Prague, Czech Republic, immediately preceding the annual meeting of the Association for
Computational Linguistics, which was hosted by Charles University. This was the second time this
workshop had been held, following the first workshop at the 2006 HLT-NAACL conference. But its
ancestry can be traced back farther to the ACL 2005 Workshop on Building and Using Parallel Texts
(whenwestartedourevaluationcampaignonEuropeanlanguages),andeventheACL2001Workshop
on Data-Driven Machine Translation (which was the first ACL workshop mostly directed at statistical
machinetranslation).
Over the last years, interest in statistical machine translation has been risen dramatically. We received
an overwhelming number of full paper submission for a one-day workshop, 38 in total. Given our
limitedcapacity,wewereonlyabletoaccept12fullpapersfororalpresentationand9papersforposter
presentation, an acceptance rate of 55%. In a second poster session, 16 additional shared task papers
werepresented. TheworkshopalsofeaturedaninvitedtalkbyJeanSenellartofSYSTRANLanguage
TranslationTechnology,Paris.
Prior to the workshop, in addition to soliciting relevant papers for review and possible presentation
we conducted a shared task that brought together machine translation systems for an evaluation on
previously unseen data. This year’s task resembled the shared tasks of previous years in many ways.
Its focus was again the translation of European languages, using a relatively large training corpus.
This year, we included a variety of manual evaluations of the MT systems’ outputs, and a variety of
automated evaluation metrics. Also, as a special challenge this year, we posed the problem of domain
adaptation.
Theresultsofthesharedtaskwereannouncedattheworkshop, andtheseproceedingsalsoincludean
overview paper for the shared task that summarizes the results, as well as provides information about
the data used and any procedures that were followed in conducting or scoring the task. In addition,
thereareshortpapersfromeachparticipatingteamthatdescribetheirunderlyingsysteminsomedetail.
We would like to thank the members of the Program Committee for their timely reviews. We also
wouldliketothanktheparticipantsofthesharedtask,theparticipantsoftheMTMarathon,whichwas
organized by the University of Edinburgh in March this year, and all the other volunteers who helped
with the manual evaluations. We also acknowledge financial support for the manual evaluation by the
EuroMatrixproject(fundedbytheEuropeanCommissionundertheFrameworkProgramme6).
ChrisCallison-Burch,PhilippKoehn,ChristofMonz,andCameronShawFordyce
Co-Organizers
iii

Organizers
Chairs:
ChrisCallison-Burch(JohnsHopkinsUniversity)
PhilippKoehn(UniversityofEdinburgh)
ChristofMonz(QueenMary,UniversityofLondon)
CameronShawFordyce(CenterfortheEvaluationofLanguageandCommunicationTechnologies)
InvitedSpeaker:
JeanSenellart(SYSTRANLanguageTranslationTechnology,Paris)
ProgramCommittee:
LarsAhrenberg(Linko¨pingUniversity)
FranciscoCasacuberta(UniversityofValencia)
ColinCherry(UniversityofAlberta)
StephenClark(OxfordUniversity)
BrookeCowan(MassachusettsInstituteofTechnology)
MonaDiab(ColumbiaUniversity)
ChrisDyer(UniversityofMaryland)
AndreasEisele(UniversitySaarbru¨cken)
MarcelloFederico(ITC-IRST)
GeorgeFoster(CanadaNationalResearchCouncil)
AlexFraser(ISI/UniversityofSouthernCalifornia)
UlrichGermann(UniversityofToronto)
RebeccaHwa(UniversityofPittsburgh)
KevinKnight(ISI/UniversityofSouthernCalifornia)
PhilippeLanglais(UniversityofMontreal)
AlonLavie(CarnegieMelonUniversity)
LoriLevin(CarnegieMellonUniversity)
DanielMarcu(ISI/UniversityofSouthernCalifornia)
BobMoore(MicrosoftResearch)
MilesOsborne(UniversityofEdinburgh)
MichelSimard(CanadaNationalResearchCouncil)
EiichiroSumita(NICT/ATR)
Jo¨rgTiedemann(UniversityofGroningen)
ChristophTillmann(IBMResearch)
DanTufis¸ (RomanianAcademy)
TaroWatanabe(NTT)
DekaiWu(HKUST)
RichardZens(RWTHAachen)
AdditionalReviewers: JoshuaAlbrecht,MarineCarpuat,HirofumiYamamoto,andKeijiYasuda.
v

Table of Contents
UsingDependencyOrderTemplatestoImproveGeneralityinTranslation
ArulMenezesandChrisQuirk.............................................................1
CCGSupertagsinFactoredStatisticalMachineTranslation
AlexandraBirch,MilesOsborneandPhilippKoehn..........................................9
IntegrationofanArabicTransliterationModuleintoaStatisticalMachineTranslationSystem
MehdiM.Kashani,EricJoanis,RolandKuhn,GeorgeFosterandFredPopowich..............17
ExploringDifferentRepresentationalUnitsinEnglish-to-TurkishStatisticalMachineTranslation
KemalOflazerandIlknurDurgarEl-Kahlout...............................................25
CanWeTranslateLetters?
DavidVilar,Jan-ThorstenPeterandHermannNey..........................................33
ADependencyTreeletStringCorrespondenceModelforStatisticalMachineTranslation
DeyiXiong,QunLiuandShouxunLin....................................................40
WordErrorRates: DecompositionoverPOSclassesandApplicationsforErrorAnalysis
MajaPopovicandHermannNey..........................................................48
Speech-InputMulti-TargetMachineTranslation
AliciaPe´rez,M.TeresaGonza´lez,M.Ine´sTorresandFranciscoCasacuberta..................56
Meta-StructureTransformationModelforStatisticalMachineTranslation
JiadongSun,TiejunZhaoandHuashenLiang..............................................64
TrainingNon-ParametricFeaturesforStatisticalMachineTranslation
PatrickNguyen,MilindMahajanandXiaodongHe.........................................72
UsingWord-DependentTransitionModelsinHMM-BasedWordAlignmentforStatisticalMachineTrans-
lation
XiaodongHe ........................................................................... 80
EfficientHandlingofN-gramLanguageModelsforStatisticalMachineTranslation
MarcelloFedericoandMauroCettolo.....................................................88
HumanEvaluationofMachineTranslationThroughBinarySystemComparisons
DavidVilar,GregorLeusch,HermannNeyandRafaelE.Banchs ............................ 96
LabelledDependenciesinMachineTranslationEvaluation
KarolinaOwczarzak,JosefvanGenabithandAndyWay...................................104
AnIteratively-TrainedSegmentation-FreePhraseTranslationModelforStatisticalMachineTranslation
RobertMooreandChrisQuirk .......................................................... 112
vii
UsingParaphrasesforParameterTuninginStatisticalMachineTranslation
NitinMadnani,NecipFazilAyan,PhilipResnikandBonnieDorr...........................120
Mixture-ModelAdaptationforSMT
GeorgeFosterandRolandKuhn.........................................................128
(Meta-)EvaluationofMachineTranslation
ChrisCallison-Burch,CameronFordyce,PhilippKoehn,ChristofMonzandJoshSchroeder...136
Context-awareDiscriminativePhraseSelectionforStatisticalMachineTranslation
Jesu´sGime´nezandLlu´ısMa`rquez.......................................................159
Ngram-BasedStatisticalMachineTranslationEnhancedwithMultipleWeightedReorderingHypotheses
MartaR.Costa-jussia`,JosepM.Crego,PatrikLambert,MaximKhalilov,Jose´ A.R.Fonollosa,Jose´
B.MarioandRafaelE.Banchs...............................................................167
Analysis of Statistical and Morphological Classes to Generate Weigthed Reordering Hypotheses on a
StatisticalMachineTranslationSystem
MartaR.Costa-jussia` andJose´ A.R.Fonollosa ........................................... 171
DomainAdaptationinStatisticalMachineTranslationwithMixtureModelling
JorgeCiveraandAlfonsJuan............................................................177
GettingtoKnowMoses: InitialExperimentsonGerman-EnglishFactoredTranslation
MariaHolmqvist,SaraStymneandLarsAhrenberg ....................................... 181
NRC’sPORTAGESystemforWMT2007
NicolaUeffing,MichelSimard,SamuelLarkinandHowardJohnson........................185
BuildingaStatisticalMachineTranslationSystemforFrenchUsingtheEuroparlCorpus
HolgerSchwenk.......................................................................189
Multi-EngineMachineTranslationwithanOpen-SourceSMTDecoder
YuChen,AndreasEisele,ChristianFedermann,EvaHasler,MichaelJellinghausand
SilkeTheison..........................................................................193
TheISLPhrase-BasedMTSystemforthe2007ACLWorkshoponStatisticalMachineTranslation
MatthiasPaulik,KayRottmann,JanNiehues,SiljaHildebrandandStephanVogel............197
Rule-BasedTranslationwithStatisticalPhrase-BasedPost-Editing
MichelSimard,NicolaUeffing,PierreIsabelleandRolandKuhn ........................... 203
The”NoisierChannel”: TranslationfromMorphologicallyComplexLanguages
ChristopherJ.Dyer.....................................................................207
UCBSystemDescriptionfortheWMT2007SharedTask
PreslavNakovandMartiHearst.........................................................212
viii
TheSyntaxAugmentedMT(SAMT)SystemattheSharedTaskforthe2007ACLWorkshoponStatistical
MachineTranslation
AndreasZollmann,AshishVenugopal,MatthiasPaulikandStephanVogel...................216
StatisticalPost-EditingonSYSTRAN’sRule-BasedTranslationSystem
Lo¨ıcDugast,JeanSenellartandPhilippKoehn............................................220
ExperimentsinDomainAdaptationforStatisticalMachineTranslation
PhilippKoehnandJoshSchroeder.......................................................224
METEOR: An Automatic Metric for MT Evaluation with High Levels of Correlation with Human Judg-
ments
AlonLavieandAbhayaAgarwal ........................................................ 228
English-to-CzechFactoredMachineTranslation
OndˇrejBojar...........................................................................232
SentenceLevelMachineTranslationEvaluationasaRanking
YangYe,MingZhouandChin-YewLin..................................................240
LocalizationofDifficult-to-TranslatePhrases
BehrangMohitandRebeccaHwa........................................................248
LinguisticFeaturesforAutomaticEvaluationofHeterogenousMTSystems
Jesu´sGime´nezandLlu´ısMa`rquez.......................................................256
ix