eduJi
yi
gChe
page1Ji
yi
gChe
U
iversityofPe
sylva
ia
HomeAddressDepartme
tofComputera
dI
formatio
Scie
ce32Marvi
La
e
3330Wal
utStreetLevi
eHallCISDeptPiscatawayNJ08854PhiladelphiaPA19104Cellpho
e7326687728Office2155737736edu
httpwwwcisupe
eduji
yi
g
Educatio
U
iversityofPe
sylva
iaPhiladelphiaPAUSAPhDca
didate2001PresDepartme
tofComputera
dI
formatio
Scie
ce
Dissertatio
TowardsHighperforma
ceWordSe
seDisambiguatio
byCombi
i
gRichLi
guisticK
owledgea
dMachi
eLear
i
gApproachestobedefe
dedi
July2006
AdvisorMarthaSPalmer
CommitteeMembersJoshiKAravi
dClaireCardieexter
alexami
er
MitchPMarcuschairLyleHU
gar
U
iversityofPe
sylva
iaPhiladelphiaPAUSAMS20002001Departme
tofComputera
dI
formatio
Scie
ce
Tsi
ghuaU
iversityBeiji
gChi
aME19982000BS19941998Departme
tofComputerScie
cea
dTech
ology
ResearchI
terests
Machi
elear
i
ga
dfeaturee
gi
eeri
gfor
aturalla
guageprocessi
gNLPAutomaticwordse
sedisambiguatio
clusteri
gsema
ticallycohere
twordsa
dautomaticacquisitio
oflargescalesema
tictaxo
omiesNLPapplicatio
stoi
formatio
retrievali
formatio
extractio
machi
etra
slatio
a
dbioi
formatics
ResearchExperie
ce
Departme
tofComputera
dI
formatio
Scie
ceU
iversityofPe
sylva
ia
PhDCa
didate20012006Highperforma
cesupervisedwordse
sedisambiguatio
WSDthroughcombi
i
gli
guisticallymotivatedfeaturesa
dasmoothedMaximumE
tropyMaxE
tmodelThesystemachievedhigheraccuracytha
previousbestsystemso
theSENSEVAL2E
glishverbdataU
superviseda
dactivelear
i
gmethodsforWSDEMclusteri
gforChi
eseverbse
sesa
dactivelear
i
gforE
glishverbse
sesClusteri
gbasedfeatureselectio
forWSDNou
clusteri
ga
dsemiautomaticallycreated
ou
taxo
omiesusedforsema
ticfeaturesforWSDNomi
ale
titydetectio
fortheChi
eseAutomaticCo
te
tExtractio
CACEprojectsummer2003summer2004Boosti
ga
dTAGTreeAdjoi
i
gGrammarSupertaggi
gfortemplaterelatio
detectio
asubtaskoftheMUC7i
formatio
extractio
taskfall2001Departme
tofComputerScie
ceTech
ologyTsi
ghuaU
iversityBeiji
gChi
a
MasterStude
tSe
iorCollegeStude
t19972000Visualizatio
dime
sio
reductio
a
dclassificatio
algorithmsforChi
esecharacterrecog
itio
Aclassificatio
algorithmbasedo
Mahala
obisdista
cea
ddime
sio
reductio
fordisti
guishi
gwellsimilarha
dwritte
Chi
esecharacters
fHo
orsa
dAwards
Graudatestude
tresearchfellowshipfromtheDepartme
tofComputera
dI
formatio
Scie
ceU
iversityofPe
sylva
iaSept2000pres
Tsi
ghuaMotorolaOutsta
di
gStude
tScholarshiptop3amo
gover50graduatestude
tsi
theDepartme
tofComputerScie
cea
dTech
ologyTsi
r