ANewsorie
tedChi
eseE
glishMachi
eTra
slatio
System
Qu
LiuBaobaoCha
gWeido
gZha
Qia
Zhou
ofComputatio
alLi
guisticsPeki
gU
iversityI
stituteofComputi
gTech
ologyChi
eseAcademyofScie
cesStateKeyLaboratoryofI
tellige
tTech
ologya
dSystemTsi
ghuaU
iversityliuqu
ictacc
chbbzhwdpkueduc
zhouqs1000ecstsi
ghuaeduc

I
stitute
AbstractThispaperi
troducesaNewsorie
tedChi
eseE
glishMachi
eTra
slatio
systemFirstlyge
erali
formatio
ofthesystemisgive
Seco
dlythearchitectureofthesystemthemicroe
gi
epipeli
eisi
troducedThirdlytheseveralmai
microe
gi
esi
thesystemarediscussedi
cludi
gmicroe
gi
esforChi
eselexicala
alysisexamplebasedmicroe
gi
ea
drulebasedmicroe
gi
eFourthlyli
guisticsresourcesusedi
thissystemaredescribedFi
allywewillgivetheco
clusio
a
dthefutureworksKeywordsMultiE
gi
eMachi
eTra
slatio
Chi
eseI
formatio
Processi
g
面向新闻领域的汉英机器翻译系统1
刘群常宝宝詹卫东周强
北京大学计算语言学研究所中国科学院计算技术研究所清华大学智能技术与系统国家重点实验室


liuqu
ictacc
chbbzhwdpkueduc
zhouqs1000ecstsi
ghuaeduc

摘要本文介绍了一个面向新闻领域的汉英机器翻译系统。首先介绍了系统的总体情况。其次，介绍了该系统的总体结构微引擎流水线。再次，讨论了系统中几个主要的微引擎了，包括词法分析微引擎，基于实例的翻译微引擎和基于规则的翻译微引擎。然后，介绍了系统中所使用的语言资源。最后给出了结论和下一步的工作计划。关键词：多引擎机器翻译，中文信息处理多引擎机器翻译，多引擎机器翻译
1I
troductio
TheNewsorie
tedChi
eseE
glishMachi
eTra
slatio
Systemisaprojectsupportedbythe
atio
alkeyfu
dame
talresearchschemaI
stituteofComputatio
alLi
guisticsofPeki
gU
iversity
fI
stituteofComputi
gTech
ologyofChi
eseAcademyofScie
cesa
dStateKeyLaboratoryofI
tellige
tTech
ologya
dSystemofTsi
ghuaU
iversityu
dertakethisprojectTheleaderoftheprojectisProfYuShiwe
Theprojectstartedo
Ja
uary2000I
thissystemweadoptamultistrategyapproachwhichwecallitamicroe
gi
epipeli
eapproachNowthefirstversio
ofthesystemiscompletedThesystemhasadictio
aryofabout50000Chi
esewordsi
thisversio
Itca
tra
slate1000wordspermi
uteo
acomputerwitha
I
telPe
tiumIII500CPU
2Multimicroe
gi
eApproach
Variousapproachesareusedi
differe
tMTsystemssuchasrulebasedapproachstatisticalbasedapproachexamplebasedapproachtra
sferbasedapproachi
terli
guabasedapproacha
detcEachapproachhasitsadva
tagesa
ddisadva
tagesAlotofresearchersrealizedthatusi
gahybridapproachr