I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
1
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
,
p
p
.
4
6
1
4
~
4
6
2
2
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
14
.i
6
.
p
p
4
6
1
4
-
4
6
2
2
4614
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Tra
nsfo
rmer and
text aug
menta
tio
n f
o
r t
o
u
rism
as
p
ect
-
ba
sed
sentimen
t
a
na
ly
sis
Sa
m
uel Sit
um
ea
ng
,
Sa
ra
h R
o
s
dia
na
T
a
m
bu
na
n,
J
ev
a
nia
,
M
a
s
t
a
wila
F
ebry
a
nti
Sim
a
njunt
a
k
,
Sa
nd
ro
Sin
a
g
a
I
n
f
o
r
mat
i
o
n
S
y
st
e
m
S
t
u
d
y
P
r
o
g
r
a
m,
F
a
c
u
l
t
y
o
f
I
n
f
o
r
ma
t
i
c
s
a
n
d
El
e
c
t
r
i
c
a
l
E
n
g
i
n
e
e
r
i
n
g
,
I
n
st
i
t
u
t
T
e
k
n
o
l
o
g
i
D
e
l
,
T
o
b
a
,
I
n
d
o
n
e
si
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Au
g
2
4
,
2
0
2
4
R
ev
is
ed
Oct
2
8
,
2
0
2
5
Acc
ep
ted
No
v
8
,
2
0
2
5
Th
e
3
6
.
9
8
%
g
r
o
wth
in
t
h
e
q
u
a
n
t
it
y
o
f
e
lec
tro
n
ic
w
o
rd
o
f
m
o
u
t
h
(e
-
WOM
)
o
v
e
r
th
e
p
a
st
fi
v
e
y
e
a
rs
p
re
se
n
ts
o
p
p
o
rt
u
n
it
ies
fo
r
th
e
to
u
rism
in
d
u
stry
to
u
n
d
e
rsta
n
d
t
o
u
rists'
n
e
e
d
s
a
n
d
d
e
sire
s
b
e
tt
e
r
wh
e
n
a
n
a
l
y
z
e
d
e
ffe
c
ti
v
e
ly
.
As
p
e
c
t
-
b
a
se
d
se
n
ti
m
e
n
t
a
n
a
ly
sis
(ABSA)
is
p
ro
p
o
se
d
a
s
a
so
lu
ti
o
n
,
a
s
it
c
a
n
id
e
n
ti
f
y
t
h
e
se
n
ti
m
e
n
t
a
t
a
m
o
re
d
e
tailed
a
sp
e
c
t
lev
e
l.
P
rio
r
re
se
a
rc
h
re
v
e
a
led
two
issu
e
s
in
ABSA:
imb
a
la
n
c
e
d
d
a
tas
e
ts
a
n
d
p
o
o
r
p
e
rfo
r
m
a
n
c
e
in
re
p
re
se
n
ti
n
g
imp
li
c
it
a
sp
e
c
ts
a
n
d
o
p
in
i
o
n
s.
T
h
e
a
u
th
o
rs
p
r
o
p
o
se
d
a
c
o
m
b
in
a
ti
o
n
o
f
t
h
e
b
id
irec
ti
o
n
a
l
a
n
d
a
u
to
-
re
g
re
ss
iv
e
tran
sfo
rm
e
r
(BART)
a
n
d
b
id
irec
ti
o
n
a
l
e
n
c
o
d
e
r
re
p
re
se
n
tatio
n
s
fr
o
m
tran
sf
o
rm
e
rs
(BERT
)
m
o
d
e
ls.
Lev
e
ra
g
in
g
BART
c
a
p
a
b
il
it
y
in
m
o
d
e
li
n
g
c
o
n
tex
t
a
n
d
BERT
e
x
p
e
rti
se
in
m
o
d
e
li
n
g
tex
t
se
m
a
n
ti
c
s
a
n
d
n
u
a
n
c
e
s,
th
e
a
u
th
o
r
p
r
o
p
o
se
d
a
n
AB
S
A
m
o
d
e
l
th
a
t
c
o
m
b
i
n
e
s
BART
a
n
d
B
ERT
u
si
n
g
th
e
e
n
se
m
b
le
m
e
t
h
o
d
.
T
h
e
e
x
p
e
rime
n
tal
re
su
lt
s
re
v
e
a
l
t
h
a
t
c
o
m
b
i
n
in
g
th
e
se
m
o
d
e
ls
sig
n
ifi
c
a
n
tl
y
e
n
h
a
n
c
e
s
th
e
p
e
rfo
rm
a
n
c
e
o
f
th
e
ABSA
m
o
d
e
l,
with
a
n
F
1
-
sc
o
re
re
a
c
h
in
g
7
0
%
.
F
u
rt
h
e
rm
o
re
,
te
x
t
a
u
g
m
e
n
tat
io
n
a
n
d
p
re
p
r
o
c
e
ss
in
g
d
i
d
n
o
t
b
ri
n
g
imp
ro
v
e
m
e
n
ts i
n
ABSA
p
e
rfo
rm
a
n
c
e
.
K
ey
w
o
r
d
s
:
A
s
p
e
c
t
-
b
as
e
d
s
e
n
t
i
m
e
n
t
a
n
a
l
y
s
i
s
B
AR
T
B
E
R
T
Sen
tim
en
t a
n
aly
s
is
T
o
u
r
is
m
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Sam
u
el
Sit
u
m
ea
n
g
I
n
f
o
r
m
atio
n
Sy
s
tem
Stu
d
y
Pr
o
g
r
am
,
Facu
lty
o
f
I
n
f
o
r
m
atics a
n
d
E
lectr
ical
E
n
g
in
ee
r
in
g
I
n
s
titu
t T
ek
n
o
lo
g
i D
el
St
.
Sis
in
g
am
an
g
ar
aja,
Sit
o
lu
a
m
a
,
L
ag
u
b
o
ti,
T
o
b
a
,
No
r
t
h
Su
m
atr
a
Pro
v
in
ce
,
I
n
d
o
n
esia
E
m
ail:
s
am
u
el.
s
itu
m
ea
n
g
@
d
el.
ac
.
id
1.
I
NT
RO
D
UCT
I
O
N
E
lectr
o
n
ic
wo
r
d
o
f
m
o
u
th
(e
-
W
OM
)
is
an
ac
tiv
ity
wh
er
e
e
x
p
er
ien
ce
d
co
n
s
u
m
er
s
s
h
ar
e
i
n
f
o
r
m
atio
n
an
d
r
ec
o
m
m
en
d
atio
n
s
o
n
lin
e
r
eg
ar
d
in
g
ce
r
tain
v
e
n
d
o
r
s
o
r
p
r
o
d
u
cts
[
1
]
,
[
2
]
.
T
h
e
Statis
ta
s
u
r
v
ey
r
ep
o
r
ts
th
at
T
r
ip
Ad
v
is
o
r
’
s
e
-
W
OM
witn
ess
ed
an
asto
u
n
d
in
g
3
6
.
9
8
%
g
r
o
wth
f
r
o
m
2
0
1
8
to
2
0
2
2
.
T
h
e
s
u
r
v
ey
f
u
r
th
e
r
h
ig
h
lig
h
ts
an
an
n
u
al
in
cr
ea
s
e
in
e
-
W
OM
q
u
a
n
tity
,
with
p
r
o
jectio
n
s
f
o
r
co
n
tin
u
ed
r
is
in
g
in
t
h
e
u
p
co
m
in
g
y
ea
r
s
.
T
h
e
g
r
o
wth
u
n
d
o
u
b
ted
ly
p
r
esen
ts
o
p
p
o
r
t
u
n
ities
f
o
r
th
e
to
u
r
is
m
in
d
u
s
tr
y
t
o
u
n
d
er
s
tan
d
th
e
to
u
r
is
ts
'
n
ee
d
s
an
d
d
esire
s
wh
en
an
aly
z
ed
ef
f
ec
tiv
ely
[
3
]
.
On
e
f
o
r
m
o
f
e
-
W
OM
an
aly
s
i
s
is
s
en
tim
en
t
an
aly
s
is
,
a
n
at
u
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
(
NL
P)
task
to
ev
alu
ate
th
e
tex
t'
s
s
en
tim
en
t
[
4
]
.
Ho
wev
er
,
s
en
tim
en
t
an
aly
s
is
ca
n
n
o
t
b
e
u
s
ed
to
f
in
d
o
u
t
th
e
s
p
ec
if
ic
asp
ec
ts
r
ev
iewe
d
b
y
v
is
ito
r
s
.
T
h
er
ef
o
r
e,
asp
ec
t
-
b
ased
s
en
tim
en
t
a
n
al
y
s
is
(
AB
SA)
is
p
r
o
p
o
s
ed
as
a
s
o
lu
tio
n
s
o
th
at
t
h
e
s
en
tim
en
t
o
f
asp
ec
ts
co
n
tain
e
d
ca
n
b
e
id
e
n
tifie
d
[
5
]
.
Fo
r
th
e
to
u
r
is
m
s
ec
to
r
,
AB
SA
is
ap
p
lied
to
u
n
d
er
s
tan
d
to
u
r
is
t sen
tim
en
t to
war
d
s
ce
r
t
ain
asp
ec
ts
o
f
a
to
u
r
is
t a
ttra
ctio
n
.
I
n
AB
SA
r
esear
ch
,
t
h
e
f
o
u
r
m
ain
f
o
c
u
s
elem
en
ts
ar
e
asp
ec
t
ca
teg
o
r
y
(
C
)
,
asp
ec
t
te
r
m
(
A
)
,
s
en
tim
en
t
p
o
lar
ity
(
S),
an
d
o
p
in
io
n
ter
m
(
O)
[
6
]
.
Fo
r
in
s
tan
ce
,
co
n
s
id
er
in
g
th
e
s
en
te
n
ce
"th
e
s
taf
f
h
er
e
is
v
er
y
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Tr
a
n
s
fo
r
mer a
n
d
text
a
u
g
men
t
a
tio
n
fo
r
to
u
r
is
m
a
s
p
ec
t
-
b
a
s
ed
s
en
timen
t a
n
a
lysi
s
(
S
a
mu
el
S
i
tu
mea
n
g
)
4615
f
r
ien
d
ly
!
"
,
th
e
r
elev
a
n
t
elem
e
n
ts
ar
e
"staf
f
"
(
A)
,
"ser
v
ice"
(
C
)
,
"q
u
ality
o
f
s
er
v
ice"
(
O)
,
an
d
"p
o
s
itiv
e"
(
S).
T
h
is
allo
ws co
m
p
an
ies to
g
ain
m
o
r
e
m
ea
n
in
g
f
u
l a
n
d
s
p
ec
if
ic
in
s
ig
h
t in
to
cu
s
to
m
er
ev
alu
ati
o
n
s
.
Ar
ian
to
an
d
B
u
d
i
[
7
]
r
e
v
ea
led
th
at
t
h
e
p
r
o
p
o
s
ed
m
o
d
el
p
e
r
f
o
r
m
a
n
ce
is
s
u
b
o
p
tim
al.
H
o
wev
er
,
th
e
s
tu
d
y
d
o
es
n
o
t
s
p
ec
if
ically
elu
cid
ate
th
e
r
ea
s
o
n
s
f
o
r
th
is
.
Po
ten
tial
ca
u
s
es
ca
n
b
e
f
o
u
n
d
in
s
tu
d
ies
[
8
]
,
[
9
]
,
wh
er
e
s
ev
er
al
is
s
u
es
ar
e
id
en
tifie
d
.
T
h
e
f
ir
s
t
is
s
u
e
is
th
at
th
e
p
r
o
p
o
s
ed
m
o
d
el
is
s
en
s
itiv
e
to
th
e
im
b
alan
ce
d
d
is
tr
ib
u
tio
n
o
f
d
ata
[
9
]
.
On
e
o
f
th
e
d
atasets
u
s
ed
is
r
estau
r
an
t
-
AC
OS,
wh
ich
h
as
a
d
ata
d
is
tr
ib
u
tio
n
h
ea
v
il
y
s
k
ewe
d
to
war
d
s
th
e
q
u
ad
r
u
p
l
e
ty
p
e
ex
p
licit
asp
ec
t
with
ex
p
licit
o
p
in
io
n
(
E
AE
O)
at
6
4
%,
co
m
p
ar
ed
to
th
e
o
th
er
th
r
ee
ty
p
es:
im
p
licit
asp
ec
t
with
ex
p
licit
o
p
in
io
n
(
I
AE
O)
,
ex
p
licit
asp
ec
t
with
im
p
lic
it
o
p
in
io
n
(
E
AI
O)
,
an
d
im
p
licit
asp
ec
t
with
im
p
licit
o
p
in
io
n
(
I
AI
O)
.
T
h
e
s
ec
o
n
d
is
s
u
e
is
th
e
m
o
d
el'
s
p
er
f
o
r
m
an
ce
o
n
q
u
ad
r
u
p
le
ty
p
es
co
n
tain
in
g
im
p
licit
a
s
p
ec
ts
o
r
im
p
licit
o
p
in
i
o
n
s
,
wh
ich
r
em
ai
n
s
u
n
s
atis
f
ac
to
r
y
.
T
h
e
m
o
d
el'
s
p
er
f
o
r
m
an
ce
o
n
th
e
E
AI
O
q
u
ad
r
u
p
le
ty
p
e
in
b
o
th
d
atasets
s
h
o
ws
n
o
tab
ly
lo
w
f
i
g
u
r
es,
n
a
m
ely
2
3
.
4
%
f
o
r
t
h
e
lap
to
p
-
AC
OS
d
ataset
an
d
2
0
%
f
o
r
th
e
r
estau
r
an
t
-
AC
OS
d
ataset.
T
h
e
I
AE
O
q
u
ad
r
u
p
le
t
y
p
e
s
h
o
ws
f
ig
u
r
es
o
f
5
2
.
7
9
%
o
n
l
ap
to
p
-
AC
OS
an
d
4
3
.
8
7
%
o
n
r
estau
r
a
n
t
-
AC
OS,
wh
ile
th
e
I
AI
O
q
u
ad
r
u
p
le
ty
p
e
s
h
o
ws
f
ig
u
r
es
o
f
2
9
.
7
9
%
o
n
l
ap
to
p
-
AC
OS
an
d
4
2
.
8
6
%
o
n
r
estau
r
an
t
-
AC
OS.
T
h
e
r
esu
lts
f
r
o
m
th
is
s
tu
d
y
in
d
icate
th
at
th
e
p
r
o
p
o
s
ed
m
o
d
el
h
as
n
o
t
y
et
b
ee
n
ab
le
to
r
e
p
r
esen
t
im
p
licit
asp
ec
ts
o
r
im
p
licit
o
p
in
io
n
s
ef
f
ec
tiv
ely
.
Hen
ce
,
a
r
o
b
u
s
t
ap
p
r
o
ac
h
to
im
b
alan
ce
d
d
ata
an
d
g
o
o
d
ch
a
r
ac
ter
is
tics
in
r
ep
r
esen
tin
g
im
p
licit
o
p
in
io
n
s
an
d
asp
ec
ts
in
r
ev
iews is
n
ec
ess
ar
y
to
ad
d
r
es
s
th
ese
is
s
u
es.
T
h
e
p
r
ev
io
u
s
ap
p
r
o
ac
h
th
at
ca
n
b
e
u
s
ed
as
an
alter
n
ativ
e
f
o
r
im
b
alan
ce
d
d
ata
p
r
o
b
le
m
s
is
d
ata
au
g
m
en
tatio
n
.
Data
a
u
g
m
e
n
tatio
n
h
elp
s
ac
h
iev
e
s
ev
er
al
g
o
als
s
u
ch
as
r
e
g
u
lar
izatio
n
,
r
ed
u
cin
g
r
ea
l
-
wo
r
l
d
d
ata,
esp
ec
ially
in
p
r
iv
ac
y
-
s
en
s
itiv
e
d
o
m
ain
s
,
m
in
im
izin
g
la
b
elin
g
ef
f
o
r
t,
b
alan
cin
g
im
b
al
an
ce
d
d
atasets
,
an
d
en
h
an
cin
g
r
esil
ien
ce
to
ad
v
er
s
ar
ial
ex
am
p
les
[
1
0
]
.
Sh
o
r
ten
an
d
Kh
o
s
h
g
o
f
taar
[
1
0
]
r
ev
ea
led
th
at
d
ata
au
g
m
en
tatio
n
ca
n
n
o
t
o
v
er
co
m
e
all
p
o
s
s
ib
le
tr
an
s
f
o
r
m
atio
n
s
an
d
elim
in
ate
all
ty
p
es
o
f
b
ias
in
th
e
d
ata.
As
an
ex
am
p
le
o
f
b
ias
in
d
ata
f
r
o
m
r
esear
ch
[
1
0
]
,
if
th
e
d
ataset
in
a
n
ews
class
if
icatio
n
task
d
o
es
n
o
t
c
o
n
tain
s
p
o
r
ts
-
r
elate
d
ar
ticles,
th
en
th
e
d
ata
au
g
m
en
tatio
n
m
eth
o
d
u
s
ed
b
a
s
ed
o
n
th
at
d
ata
s
et
will
m
o
s
t
lik
ely
n
o
t
p
r
o
d
u
ce
d
ata
r
elate
d
to
s
p
o
r
ts
ar
ticle
s
,
ev
en
if
it
is
ess
en
tial.
D
ata
au
g
m
en
tatio
n
ca
n
in
tr
o
d
u
ce
s
u
p
p
lem
en
ta
r
y
,
u
n
d
esira
b
le
b
iases
th
at
r
esu
lt
i
n
an
in
ac
cu
r
ate
r
ep
r
esen
tatio
n
o
f
th
e
e
n
tire
p
o
p
u
latio
n
.
Fo
r
in
s
tan
ce
,
lan
g
u
a
g
e
m
o
d
els
lik
e
g
en
e
r
ativ
e
p
r
e
-
tr
ain
ed
tr
an
s
f
o
r
m
er
(
GPT)
ca
n
cr
ea
te
b
iases
th
at
ar
e
th
en
t
r
an
s
f
er
r
ed
i
n
to
th
e
d
ataset
s
u
ch
as
th
e
h
ig
h
co
r
r
e
latio
n
b
etwe
en
th
e
wo
r
d
“c
r
i
m
in
al”
an
d
m
ale
id
en
tity
in
t
h
e
GPT
-
2
o
u
tp
u
t,
as
well
as “
Go
d
”
an
d
C
h
r
is
tian
ity
.
T
h
is
ca
n
ca
u
s
e
th
e
m
o
d
el
t
o
p
r
o
d
u
ce
b
iased
d
ec
is
io
n
s
[
1
1
]
.
T
h
er
e
ar
e
v
ar
io
u
s
d
ata
au
g
m
e
n
tatio
n
tech
n
iq
u
es
with
v
ar
y
i
n
g
co
m
p
lex
it
y
.
Ho
wev
er
,
co
m
p
lex
d
ata
au
g
m
en
tatio
n
tech
n
i
q
u
es
ten
d
to
b
e
in
e
f
f
icien
t.
I
n
cr
ea
s
ed
d
em
an
d
o
n
r
eso
u
r
ce
s
,
esp
ec
ially
wh
en
tr
ain
i
n
g
g
en
er
ativ
e
m
o
d
els,
is
a
n
atu
r
a
l
p
ar
t
o
f
d
ata
a
u
g
m
e
n
tatio
n
.
T
o
m
itig
ate
s
o
m
e
o
f
th
e
lim
itat
io
n
s
an
d
m
ax
im
ize
th
e
ad
v
a
n
tag
es
o
f
d
ata
au
g
m
en
tatio
n
,
t
h
e
au
t
h
o
r
s
p
r
o
p
o
s
e
im
p
r
o
v
em
en
ts
to
e
x
is
tin
g
d
ata
au
g
m
en
tatio
n
ap
p
r
o
ac
h
es
b
ased
o
n
s
u
g
g
esti
o
n
s
p
r
o
p
o
s
ed
in
[
1
2
]
.
L
o
n
g
p
r
e
et
a
l
.
[
1
2
]
s
tated
th
at
s
ev
er
a
l
d
ata
a
u
g
m
e
n
tatio
n
tech
n
iq
u
es
d
o
n
o
t
i
n
cr
ea
s
e
th
e
p
er
f
o
r
m
an
ce
o
f
p
r
e
-
tr
ain
e
d
t
r
an
s
f
o
r
m
er
m
o
d
els
s
u
ch
as
b
id
ir
ec
tio
n
al
en
c
o
d
er
r
ep
r
esen
tatio
n
s
f
r
o
m
tr
an
s
f
o
r
m
er
s
(
B
E
R
T
)
,
r
o
b
u
s
tly
o
p
tim
i
ze
d
B
E
R
T
p
r
etr
ain
in
g
ap
p
r
o
a
ch
(
R
o
B
E
R
T
a
)
,
an
d
ex
tr
a
-
lo
n
g
n
eu
r
al
n
etwo
r
k
(
X
L
Net
)
o
n
s
im
p
le
class
if
icatio
n
task
s
.
T
h
ey
h
y
p
o
th
esized
th
at
d
ata
au
g
m
en
tatio
n
tech
n
iq
u
es
wo
u
ld
o
n
l
y
b
e
u
s
ef
u
l
if
th
ey
c
o
u
ld
p
r
o
d
u
ce
n
o
v
el
lin
g
u
is
tic
p
atter
n
s
th
at
h
ad
n
e
v
er
b
ee
n
en
co
u
n
ter
e
d
b
e
f
o
r
e.
Me
an
wh
ile,
a
p
r
ev
i
o
u
s
ap
p
r
o
ac
h
as
an
alter
n
ativ
e
to
r
ep
r
e
s
en
tin
g
im
p
licit
o
p
in
io
n
s
f
o
r
m
en
tio
n
ed
asp
ec
ts
in
r
ev
iews
is
b
y
co
m
b
in
in
g
b
id
ir
ec
tio
n
al
an
d
au
to
-
r
eg
r
ess
iv
e
tr
an
s
f
o
r
m
er
(
B
AR
T
)
an
d
B
E
R
T
u
s
in
g
th
e
en
s
em
b
le
m
eth
o
d
.
L
ewis
et
a
l.
[
1
3
]
in
d
icate
s
th
at
co
m
b
in
in
g
th
ese
two
tr
an
s
f
o
r
m
er
m
o
d
els
u
s
in
g
en
s
em
b
le
m
eth
o
d
s
ig
n
if
ican
tl
y
en
h
a
n
ce
s
th
e
p
e
r
f
o
r
m
an
ce
o
f
tr
an
s
f
o
r
m
er
m
o
d
els
to
p
r
o
v
i
d
e
m
o
r
e
r
o
b
u
s
t
an
d
ef
f
icien
t
s
o
lu
tio
n
s
f
o
r
v
ar
i
o
u
s
NL
P
task
s
.
B
y
in
teg
r
atin
g
B
AR
T
'
s
s
tr
en
g
th
s
in
m
o
d
elin
g
co
n
tex
t
an
d
B
E
R
T
's
s
tr
en
g
th
s
in
u
n
d
er
s
tan
d
in
g
tex
t'
s
em
an
tics
an
d
n
u
a
n
ce
s
,
t
h
is
ap
p
r
o
a
ch
ca
n
p
r
o
v
i
d
e
m
o
r
e
co
m
p
r
e
h
en
s
iv
e
p
er
f
o
r
m
an
ce
ac
r
o
s
s
v
ar
io
u
s
NL
P
task
s
.
T
h
e
en
s
em
b
le
[
1
4
]
m
eth
o
d
'
s
u
s
e
en
ab
les
th
e
co
n
c
u
r
r
en
t
lev
er
a
g
in
g
o
f
b
o
th
m
o
d
els'
s
tr
en
g
th
s
,
th
u
s
d
im
in
is
h
in
g
ea
ch
m
o
d
el'
s
s
h
o
r
tco
m
in
g
s
an
d
im
p
r
o
v
in
g
th
e
p
er
f
o
r
m
a
n
ce
o
f
th
e
r
esu
ltin
g
m
o
d
el
[
1
5
]
.
A
f
ac
to
r
th
at
m
a
y
co
n
tr
ib
u
te
t
o
th
e
m
o
d
el'
s
p
er
f
o
r
m
an
ce
is
th
e
d
ataset
q
u
ality
[
1
6
]
.
I
n
NL
P
task
s
,
it
'
s
n
ec
ess
ar
y
to
ac
cu
r
ately
r
ep
r
e
s
en
t
th
e
ch
ar
ac
ter
s
o
r
wo
r
d
s
in
ea
ch
s
en
ten
ce
to
ac
h
iev
e
h
ig
h
-
q
u
ality
d
ata.
C
o
n
v
er
s
ely
,
p
o
o
r
d
ata
q
u
ality
m
ay
af
f
ec
t
th
e
m
o
d
el'
s
p
er
f
o
r
m
an
ce
.
R
ep
r
esen
tin
g
th
ese
u
n
its
p
r
esen
ts
d
iv
er
s
e
ch
allen
g
es
d
ep
e
n
d
in
g
o
n
th
e
lan
g
u
a
g
e
b
ei
n
g
p
r
o
ce
s
s
ed
an
d
wr
itin
g
s
y
s
tem
s
[
1
7
]
.
T
h
is
k
n
o
wn
as
tex
t
p
r
ep
r
o
ce
s
s
in
g
[
1
8
]
,
n
ee
d
s
to
b
e
ap
p
lied
to
in
v
esti
g
ate
th
e
d
ataset
ch
ar
ac
ter
is
tic
s
th
at
alig
n
with
th
e
p
r
o
p
o
s
e
d
m
o
d
el
in
h
an
d
lin
g
th
e
AB
SA c
ase
an
d
d
eter
m
in
i
n
g
its
s
ig
n
if
ican
ce
.
Hen
ce
,
th
is
s
tu
d
y
o
b
jectiv
e
was
to
ad
d
r
ess
th
e
id
en
tifie
d
s
h
o
r
tco
m
in
g
s
b
y
lev
er
a
g
in
g
tr
an
s
f
o
r
m
er
lan
g
u
ag
e
m
o
d
els.
T
o
th
e
b
est
o
f
o
u
r
k
n
o
wled
g
e,
t
h
is
in
itiativ
e
m
ar
k
s
a
n
o
v
el
ap
p
r
o
ac
h
to
in
teg
r
atin
g
B
AR
T
an
d
B
E
R
T
u
s
in
g
an
e
n
s
em
b
l
e
m
eth
o
d
,
co
m
p
lem
en
ted
b
y
tex
t
au
g
m
en
tatio
n
an
d
p
r
e
p
r
o
ce
s
s
in
g
,
to
ad
d
r
ess
AB
SA c
h
allen
g
es.
T
h
e
k
ey
co
n
tr
ib
u
tio
n
s
o
f
th
is
p
ap
er
ar
e
as
f
o
llo
ws:
i)
Dev
elo
p
m
en
t
o
f
an
ef
f
ec
tiv
e
tr
an
s
f
o
r
m
er
m
o
d
el
f
o
r
r
ep
r
e
s
en
tin
g
im
p
licit
asp
ec
ts
an
d
o
p
in
io
n
s
f
r
o
m
r
ev
iew
tex
ts
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
4
6
1
4
-
4
6
2
2
4616
ii)
Sig
n
if
ican
ce
o
f
t
h
e
tex
t a
u
g
m
e
n
tatio
n
im
p
ac
t o
n
th
e
B
AR
T
+BERT
en
s
em
b
le
m
o
d
el.
iii)
Sig
n
if
ican
ce
o
f
t
h
e
tex
t p
r
ep
r
o
ce
s
s
in
g
im
p
ac
t o
n
th
e
B
AR
T
+BERT
en
s
em
b
le
m
o
d
el.
T
h
e
p
ap
e
r
is
s
tr
u
ctu
r
ed
i
n
th
e
f
o
llo
win
g
way
:
s
ec
tio
n
2
d
escr
ib
es
th
e
p
r
o
p
o
s
ed
m
o
d
els
a
n
d
d
etails
th
e
ex
p
er
im
e
n
tal
s
etu
p
.
Sectio
n
3
a
n
aly
ze
s
th
e
r
esu
lts
an
d
ev
alu
ates
th
e
m
o
d
els’
p
er
f
o
r
m
an
ce
.
Fin
ally
,
s
ec
tio
n
4
co
n
clu
d
es with
a
d
is
cu
s
s
io
n
o
f
f
u
t
u
r
e
wo
r
k
.
2.
M
E
T
H
O
D
T
h
is
s
tu
d
y
p
r
o
p
o
s
es
a
n
o
v
el
AB
SA
m
o
d
el
th
at
in
teg
r
ates
th
e
s
tr
en
g
th
s
o
f
two
p
o
wer
f
u
l
t
r
an
s
f
o
r
m
er
ar
ch
itectu
r
es:
B
AR
T
[
1
3
]
a
n
d
B
E
R
T
[
1
9
]
.
T
h
e
y
ar
e
p
r
e
-
t
r
ain
ed
lan
g
u
ag
e
m
o
d
els
(
PL
Ms)
with
im
p
r
ess
iv
e
ca
p
ab
ilit
ies
th
at
ar
e
ap
p
lic
ab
le
to
v
a
r
io
u
s
NL
P
task
s
.
T
h
e
p
r
o
p
o
s
ed
AB
SA
m
o
d
el
lev
er
a
g
es
th
e
lar
g
e
lan
g
u
a
g
e
m
o
d
el
s
(
L
L
M
s
)
:
B
A
R
T
L
AR
GE
an
d
B
E
R
T
B
ASE
u
n
ca
s
ed
.
I
n
itially
,
b
o
t
h
wer
e
d
esig
n
ed
as
g
en
er
al
-
p
u
r
p
o
s
e
tr
an
s
f
o
r
m
er
s
,
ap
p
licab
le
to
a
b
r
o
a
d
r
a
n
g
e
o
f
NL
P
task
s
.
T
h
e
y
s
er
v
e
d
is
tin
ct
f
u
n
ctio
n
s
.
B
E
R
T
is
s
u
itab
le
f
o
r
NL
P
task
s
th
at
r
eq
u
ir
e
a
d
ee
p
u
n
d
er
s
tan
d
i
n
g
o
f
s
em
an
tics
an
d
lan
g
u
ag
e
c
o
n
tex
t,
s
u
c
h
as
te
x
t
class
if
icatio
n
[
2
0
]
a
n
d
n
am
ed
en
tity
r
ec
o
g
n
itio
n
[
2
1
]
.
On
t
h
e
o
th
e
r
h
a
n
d
,
B
AR
T
is
id
ea
l
f
o
r
NL
P
task
s
th
at
in
v
o
lv
e
tex
t
g
e
n
er
atio
n
,
s
u
ch
as
tex
t
s
u
m
m
ar
izatio
n
[
2
2
]
,
q
u
esti
o
n
g
en
er
atio
n
[
2
3
]
,
cr
ea
t
iv
e
tex
t
g
en
er
atio
n
[
2
4
]
,
a
n
d
m
ac
h
in
e
tr
a
n
s
latio
n
[
1
3
]
.
T
h
ese
m
o
d
els
h
av
e
b
ee
n
tr
ain
ed
o
n
m
ass
iv
e
d
atasets
,
en
ab
lin
g
th
em
t
o
ca
p
tu
r
e
th
e
h
u
m
an
lan
g
u
a
g
e
n
u
an
ce
s
.
T
h
is
s
tu
d
y
in
tr
o
d
u
ce
d
m
o
d
if
ic
atio
n
s
to
th
e
m
o
d
el
’
s
f
in
al
(
h
e
ad
)
lay
er
b
y
ap
p
en
d
in
g
th
e
lin
ea
r
lay
er
to
ea
ch
p
r
e
-
tr
ain
e
d
m
o
d
el.
I
n
s
tead
o
f
a
s
in
g
le
o
u
tp
u
t
n
eu
r
o
n
f
o
r
s
en
tim
en
t
class
if
icatio
n
(
ty
p
ical
in
s
in
g
le
-
lab
el
task
s
)
,
th
e
h
ea
d
is
m
o
d
if
ied
to
h
a
v
e
m
u
ltip
le
o
u
t
p
u
t
n
e
u
r
o
n
s
,
o
n
e
f
o
r
ea
ch
p
o
s
s
ib
le
s
en
tim
en
t
lab
el
p
er
asp
ec
t.
T
h
e
n
ex
t
c
r
u
cial
s
tep
in
v
o
lv
es
ap
p
l
y
in
g
f
in
e
-
tu
n
i
n
g
to
p
r
ep
a
r
e
th
e
L
L
Ms
f
o
r
A
B
SA.
I
n
tr
ad
itio
n
al
s
en
tim
en
t
an
aly
s
is
,
th
e
task
is
ty
p
ically
s
in
g
le
-
lab
el
class
if
icatio
n
(
e.
g
.
,
p
o
s
itiv
e,
n
eg
a
tiv
e,
an
d
n
e
u
tr
al)
.
Ho
wev
er
,
AB
SA
ta
s
k
s
n
ec
ess
itate
th
e
m
o
d
el
to
d
is
ce
r
n
asp
ec
ts
with
in
th
e
tex
t
an
d
s
u
b
s
e
q
u
en
tly
class
if
y
th
e
s
en
tim
en
t
to
war
d
s
ea
ch
asp
ec
t.
Par
k
et
a
l.
[
2
3
]
e
n
co
m
p
a
s
s
es
s
ix
asp
ec
ts
,
n
am
ely
attr
ac
tio
n
s
,
am
en
ities
,
ac
ce
s
s
ib
ilit
y
,
im
ag
e,
p
r
ice,
a
n
d
h
u
m
an
r
eso
u
r
ce
s
,
as
p
iv
o
tal
asp
ec
ts
o
f
to
u
r
is
m
.
T
h
is
m
u
lti
-
lab
el
class
if
icatio
n
n
atu
r
e
n
ec
ess
itates
ad
ju
s
tm
en
ts
to
th
e
m
o
d
el.
Fin
e
-
tu
n
in
g
en
tails
ad
ap
tin
g
th
e
p
r
e
-
tr
ain
ed
m
o
d
els
to
th
e
AB
SA n
ee
d
s
.
T
h
e
s
u
b
s
eq
u
en
t
s
tep
in
v
o
lv
es
ca
lcu
latin
g
th
e
av
er
a
g
e
class
p
r
o
b
a
b
ilit
ies.
T
h
is
en
tails
av
e
r
ag
in
g
th
e
p
r
o
b
a
b
ilit
ies
p
r
ed
icted
b
y
th
e
m
o
d
el
ac
r
o
s
s
all
lab
els
to
d
e
ter
m
in
e
th
e
o
v
er
all
lik
elih
o
o
d
o
f
ea
ch
s
en
tim
en
t
class
.
C
o
m
p
u
tin
g
th
ese
a
v
er
a
g
es
g
ain
s
a
m
o
r
e
s
tab
le
an
d
r
eliab
le
esti
m
ate
o
f
th
e
s
en
tim
en
t
f
o
r
ea
ch
asp
ec
t,
m
itig
atin
g
th
e
im
p
ac
t
o
f
an
y
s
in
g
le
p
r
e
d
ictio
n
'
s
v
ar
iab
ilit
y
.
Su
b
s
eq
u
en
tly
,
th
e
p
r
ed
ic
tio
n
em
p
lo
y
s
th
ese
av
er
ag
ed
p
r
o
b
ab
ilit
ies
to
ascer
tain
th
e
f
in
al
asp
ec
ts
'
s
s
en
ti
m
en
t
class
if
icatio
n
.
T
h
e
m
o
d
el
ass
ig
n
s
th
e
lab
el
with
th
e
h
ig
h
est
av
er
ag
e
p
r
o
b
ab
ilit
y
,
en
s
u
r
in
g
an
ac
cu
r
ate
s
en
tim
en
t
an
aly
s
is
.
T
h
is
m
eth
o
d
lev
er
a
g
es
th
e
av
er
ag
ed
p
r
o
b
ab
ilit
ies
to
m
ak
e
well
-
f
o
u
n
d
ed
p
r
ed
ictio
n
s
,
th
er
eb
y
e
n
h
an
cin
g
th
e
m
o
d
el's
p
er
f
o
r
m
an
ce
in
m
u
lti
-
lab
el
s
en
tim
en
t c
lass
if
icatio
n
task
s
.
Fig
u
r
e
1
d
ep
icts
th
e
o
v
er
all
ar
c
h
itectu
r
e
o
f
th
e
m
o
d
el.
Fig
u
r
e
1
.
T
h
e
p
r
o
p
o
s
ed
B
AR
T
+BERT
m
o
d
el
ar
ch
itectu
r
e
T
h
is
s
tu
d
y
em
p
lo
y
ed
a
d
ata
s
et
o
f
Go
o
g
le
Ma
p
s
u
s
er
r
e
v
iews
o
f
th
e
B
o
r
o
b
u
d
u
r
T
e
m
p
le
an
d
Pra
m
b
an
an
T
em
p
le
to
u
r
is
t
attr
ac
tio
n
s
[
7
]
.
R
ev
iews
in
th
is
d
a
taset
u
s
e
I
n
d
o
n
esian
,
E
n
g
lis
h
,
an
d
s
o
m
e
u
s
e
b
o
th
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Tr
a
n
s
fo
r
mer a
n
d
text
a
u
g
men
t
a
tio
n
fo
r
to
u
r
is
m
a
s
p
ec
t
-
b
a
s
ed
s
en
timen
t a
n
a
lysi
s
(
S
a
mu
el
S
i
tu
mea
n
g
)
4617
lan
g
u
ag
es
s
im
u
ltan
eo
u
s
ly
.
T
h
er
ef
o
r
e,
la
n
g
u
a
g
e
s
tan
d
ar
d
iza
tio
n
was
ap
p
lied
to
th
e
r
e
v
ie
w
tex
t
to
E
n
g
lis
h
wh
ich
was b
ased
o
n
B
AR
T
'
s
s
u
p
er
io
r
p
er
f
o
r
m
an
ce
in
tex
t g
e
n
er
atio
n
in
E
n
g
lis
h
[
1
3
]
.
As
s
h
o
wn
in
T
a
b
le
1
,
th
e
d
ataset
is
im
b
alan
ce
d
,
wh
e
r
e
th
e
I
AE
O
q
u
ad
r
u
p
le
t
y
p
e
h
o
ld
s
m
o
r
e
p
r
o
m
in
e
n
ce
with
in
b
o
th
g
r
o
u
p
s
in
co
n
tr
ast
to
th
e
r
em
ain
in
g
q
u
ad
r
u
p
le
ty
p
es.
Ad
o
p
tin
g
an
im
b
alan
ce
d
d
ataset
ca
n
lead
to
b
iased
m
o
d
el
p
er
f
o
r
m
an
ce
,
th
e
m
o
d
el'
s
p
er
f
o
r
m
an
ce
ten
d
s
to
b
e
b
etter
o
n
l
y
f
o
r
th
e
m
ajo
r
ity
q
u
ad
r
u
p
le
ty
p
e.
T
h
er
ef
o
r
e,
t
h
is
s
tu
d
y
ap
p
lies
tex
t
au
g
m
e
n
tat
io
n
to
b
alan
ce
t
h
e
d
ataset
with
th
e
d
o
m
in
an
t
ty
p
e
am
o
u
n
t,
p
r
e
v
en
tin
g
th
e
m
o
d
el
f
r
o
m
b
ec
o
m
in
g
b
iased
to
war
d
s
th
e
m
ajo
r
ity
q
u
ad
r
u
p
le
ty
p
es.
I
n
th
is
s
tu
d
y
,
th
e
tex
t
au
g
m
e
n
tatio
n
tech
n
iq
u
es
to
b
e
em
p
lo
y
e
d
ar
e
w
o
r
d
r
ep
l
ac
em
en
t,
b
ac
k
tr
an
s
latio
n
,
an
d
r
an
d
o
m
d
eletio
n
,
co
n
s
id
er
in
g
t
h
at
th
ese
ar
e
c
o
m
m
o
n
l
y
u
s
ed
i
n
tex
t
au
g
m
en
tatio
n
[
2
5
]
.
M
o
r
eo
v
er
,
b
ased
o
n
r
esear
ch
in
[
2
6
]
,
[
2
7
]
,
it
is
s
h
o
wn
th
at
th
ese
h
av
e
a
p
o
s
itiv
e
im
p
ac
t
o
n
m
o
d
el
p
er
f
o
r
m
an
ce
.
I
n
th
e
b
ac
k
tr
an
s
latio
n
tech
n
iq
u
e,
th
e
ch
o
s
en
in
ter
m
e
d
iar
y
lan
g
u
a
g
e
is
C
h
in
ese
d
u
e
to
its
e
s
tab
lis
h
ed
u
s
e
in
p
r
io
r
s
tu
d
ies,
wh
ich
ar
e
co
n
s
id
er
ed
m
atu
r
e
[
2
8
]
a
n
d
le
v
er
ag
e
th
e
ch
ar
ac
ter
s
'
ex
ten
s
iv
e
d
iv
er
s
ity
,
ap
p
r
o
x
im
ately
4
7
,
0
3
5
,
en
ab
lin
g
th
e
ad
d
itio
n
o
f
d
iv
e
r
s
e
s
en
ten
ce
v
ar
iatio
n
s
[
2
9
]
.
T
h
e
ex
p
lo
r
ato
r
y
d
ata
a
n
aly
s
is
r
ev
ea
led
a
s
u
b
s
tan
tial
d
e
g
r
e
e
o
f
ch
a
r
ac
ter
in
c
o
n
s
is
ten
cy
with
in
th
e
d
ataset,
n
ec
ess
itat
in
g
tex
t p
r
ep
r
o
ce
s
s
in
g
to
in
v
esti
g
ate
its
im
p
ac
t o
n
m
o
d
el
p
er
f
o
r
m
a
n
ce
.
T
h
er
ef
o
r
e
,
th
is
s
tu
d
y
ap
p
lies
co
m
m
o
n
te
x
t
p
r
ep
r
o
ce
s
s
in
g
tech
n
iq
u
es
[
1
8
]
th
at
h
av
e
estab
lis
h
ed
a
b
en
e
f
icial
ef
f
ec
t
o
n
m
o
d
e
l
p
er
f
o
r
m
an
ce
in
p
r
io
r
r
esear
ch
.
T
h
ese
tech
n
iq
u
es
in
clu
d
e
ca
s
e
f
o
ld
in
g
[
3
0
]
,
s
lan
g
wo
r
d
co
n
v
er
s
io
n
[
3
1
]
,
d
ig
it
co
n
v
er
s
io
n
,
s
p
ec
ial
ch
ar
ac
ter
r
em
o
v
al
[
3
2
]
,
s
to
p
w
o
r
d
r
em
o
v
al
[
3
3
]
,
lem
m
atiza
tio
n
[
3
4
]
,
a
n
d
s
tem
m
in
g
[
3
4
]
.
C
o
n
s
eq
u
en
tly
,
th
is
s
tu
d
y
will
ex
am
in
e
th
e
im
p
ac
t
o
f
d
if
f
er
e
n
t
tex
t
au
g
m
e
n
tatio
n
an
d
p
r
ep
r
o
ce
s
s
in
g
tech
n
iq
u
es
o
n
AB
SA
m
o
d
el
p
er
f
o
r
m
a
n
ce
.
I
t
aim
s
t
o
id
e
n
tify
th
e
m
o
s
t
ef
f
ec
tiv
e
m
et
h
o
d
s
t
h
r
o
u
g
h
in
d
ep
en
d
en
t
ex
p
er
im
en
tatio
n
,
im
p
lem
e
n
tin
g
th
e
s
u
p
e
r
io
r
tex
t
a
u
g
m
en
tatio
n
tech
n
i
q
u
e
b
e
f
o
r
e
ev
alu
atin
g
ea
c
h
p
r
ep
r
o
ce
s
s
in
g
tech
n
iq
u
e
co
m
p
ar
ed
to
th
e
u
n
t
r
ea
ted
d
ataset.
T
o
en
s
u
r
e
th
e
g
e
n
er
aliza
b
ilit
y
o
f
th
e
f
in
d
i
n
g
s
an
d
m
itig
ate
p
o
ten
tial
b
iases
,
th
is
s
tu
d
y
u
tili
ze
th
e
n
o
r
m
ality
test
o
f
th
e
p
er
f
o
r
m
a
n
ce
d
ata
u
s
in
g
th
e
Sh
a
p
ir
o
-
W
ilk
test
with
in
a
5
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
f
r
am
ewo
r
k
.
Su
b
s
eq
u
en
tly
,
a
n
o
n
-
p
ar
am
etr
ic
W
ilco
x
o
n
s
ig
n
ed
-
r
a
n
k
test
will
b
e
ca
r
r
ie
d
o
u
t
to
d
eter
m
in
e
th
e
s
tatis
tical
s
ig
n
if
ican
ce
o
f
an
y
o
b
s
er
v
ed
d
if
f
er
e
n
ce
s
in
p
er
f
o
r
m
an
ce
b
etwe
en
th
e
m
o
d
els ac
r
o
s
s
th
e
f
o
ld
s
.
Giv
en
th
e
s
ig
n
if
ican
t
r
o
le
o
f
h
y
p
er
p
a
r
am
eter
s
in
m
o
d
el
o
p
tim
izatio
n
,
th
is
s
tu
d
y
d
r
aws
u
p
o
n
in
s
ig
h
ts
f
r
o
m
p
r
ev
io
u
s
r
esear
c
h
[
3
5
]
–
[
3
7
]
to
d
ef
i
n
e
th
e
h
y
p
er
p
a
r
am
eter
s
ettin
g
s
.
Ap
ar
t
f
r
o
m
ad
o
p
tin
g
s
ev
e
r
al
h
y
p
er
p
ar
am
eter
s
'
s
ettin
g
s
r
ef
er
r
ed
to
p
r
io
r
r
esear
ch
,
th
e
au
th
o
r
s
also
im
p
lem
en
ted
co
n
tr
o
lled
tr
ials
an
d
an
aly
ze
d
th
e
ex
p
er
im
en
tal
r
esu
lts
to
o
p
tim
ize
th
e
h
y
p
er
p
ar
am
eter
s
'
s
ettin
g
s
.
T
ab
le
2
lis
ts
th
e
o
p
tim
al
h
y
p
er
p
ar
am
eter
s
.
T
ab
le
1
.
Nu
m
b
er
o
f
q
u
ad
r
u
p
le
ty
p
es
p
er
r
ev
iew
g
r
o
u
p
R
e
v
i
e
w
g
r
o
u
p
Q
u
a
d
r
u
p
l
e
t
y
p
e
EA
EO
EA
I
O
I
A
EO
I
A
I
O
B
o
r
o
b
u
d
u
r
t
e
m
p
l
e
1
,
1
8
6
4
9
3
2
,
1
0
2
2
2
4
P
r
a
mb
a
n
a
n
t
e
m
p
l
e
1
,
2
4
6
6
7
9
2
,
3
0
2
3
0
9
T
ab
le
2
.
Op
tim
al
h
y
p
er
p
ar
am
e
ter
s
ettin
g
s
f
o
r
B
AR
T
-
B
E
R
T
P
a
r
a
me
t
e
r
V
a
l
u
e
s
Ep
o
c
h
10
B
a
t
c
h
si
z
e
32
Le
a
r
n
i
n
g
r
a
t
e
3
×
10
-
5
W
e
ev
alu
ate
m
o
d
el
p
er
f
o
r
m
a
n
ce
u
s
in
g
th
e
F1
-
s
co
r
e,
wh
ic
h
is
th
e
h
ar
m
o
n
ic
m
ea
n
o
f
p
r
ec
is
io
n
an
d
r
ec
all.
Pre
cisi
o
n
m
ea
s
u
r
es
th
e
r
atio
o
f
ac
cu
r
ately
p
r
e
d
icted
asp
ec
t
o
r
o
p
in
io
n
ter
m
s
o
u
t
o
f
all
th
e
p
r
e
d
icted
ter
m
s
.
I
n
co
m
p
ar
is
o
n
,
r
ec
all
is
d
ef
in
ed
as
th
e
r
atio
o
f
co
r
r
ec
t
ly
p
r
ed
icted
asp
ec
t
ter
m
s
o
r
o
p
in
io
n
ter
m
s
to
th
e
to
tal
n
u
m
b
er
o
f
asp
ec
t
ter
m
s
o
r
o
p
in
io
n
ter
m
s
in
th
e
d
ataset.
W
e
al
s
o
u
tili
ze
th
e
f
lesch
r
ea
d
in
g
ea
s
e
(
FR
E
)
[
3
8
]
an
d
th
e
g
u
n
n
i
n
g
f
o
g
in
d
e
x
(
FOG)
[
3
9
]
t
o
ass
ess
th
e
r
ea
d
ab
ilit
y
an
d
co
m
p
lex
ity
o
f
t
h
e
tex
tu
al
d
ata.
T
h
e
FR
E
ass
ig
n
s
a
s
co
r
e
f
r
o
m
0
t
o
1
0
0
,
wh
er
e
h
ig
h
e
r
n
u
m
b
e
r
s
m
ea
n
th
e
te
x
t
is
ea
s
ier
to
r
ea
d
.
T
h
e
FOG
in
d
e
x
,
co
n
v
er
s
ely
,
esti
m
ates
th
e
y
ea
r
s
o
f
s
ch
o
o
lin
g
n
ee
d
ed
t
o
u
n
d
e
r
s
tan
d
th
e
tex
t
o
n
a
f
ir
s
t
r
ea
d
i
n
g
.
A
h
ig
h
er
FOG
s
co
r
e
m
ea
n
s
th
e
tex
t is m
o
r
e
c
o
m
p
lex
,
with
a
s
co
r
e
ab
o
v
e
1
4
g
en
er
ally
c
o
n
s
id
er
ed
d
if
f
icu
lt.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
p
r
o
p
o
s
ed
m
o
d
els'
ef
f
ec
tiv
en
ess
was
ass
e
s
s
ed
th
r
o
u
g
h
v
ar
io
u
s
ex
p
er
im
en
ts
to
ev
al
u
ate
th
eir
ca
p
ab
ilit
ies.
T
h
e
f
o
llo
win
g
s
u
b
s
ec
tio
n
s
s
h
o
w
an
d
an
al
y
ze
th
e
m
o
d
el
p
e
r
f
o
r
m
an
ce
ev
al
u
atio
n
f
o
r
AB
SA.
3
.
1
.
Co
m
pa
ra
t
iv
e
a
na
ly
s
is
o
f
AB
SA
m
o
del
T
ab
le
3
d
is
p
lay
s
th
e
m
o
d
el
p
er
f
o
r
m
a
n
ce
b
en
c
h
m
ar
k
in
g
.
T
h
e
r
esu
lts
s
h
o
w
th
at
th
e
p
r
o
p
o
s
ed
m
o
d
el
d
em
o
n
s
tr
ates
a
g
o
o
d
ca
p
ab
ilit
y
to
id
en
tify
th
e
asp
ec
t
ter
m
s
an
d
ev
alu
ate
s
en
tim
en
t
p
o
lar
ity
.
T
h
e
ex
p
er
im
en
t
r
esu
lts
s
h
o
wca
s
ed
in
T
ab
le
3
in
d
icate
th
at
th
e
B
AR
T
+BER
T
en
s
em
b
le
m
o
d
el
ac
co
m
p
lis
h
ed
s
u
p
er
io
r
r
esu
lts
co
m
p
ar
ed
to
r
ec
e
n
t
ap
p
r
o
ac
h
e
s
:
B
E
R
T
[
1
9
]
,
k
-
n
ea
r
est
n
eig
h
b
o
r
s
(
KNN)
[
4
0
]
,
lin
ea
r
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
4
6
1
4
-
4
6
2
2
4618
(
SVM)
[
4
1
]
,
r
ad
ial
b
asis
f
u
n
ctio
n
(
R
B
F)
-
SVM
[
4
2
]
,
d
e
cisi
o
n
tr
ee
[
4
3
]
,
r
an
d
o
m
f
o
r
est
[
4
4
]
,
m
u
ltil
ay
er
p
er
ce
p
tr
o
n
(
ML
P)
[
4
5
]
,
Ad
aBo
o
s
t
[
4
6
]
,
n
aiv
e
B
ay
es
[
4
7
]
,
a
n
d
q
u
a
d
r
atic
d
is
cr
im
in
an
t
an
al
y
s
is
(
QDA)
[
4
8
]
.
T
ab
le
3
.
Per
f
o
r
m
an
ce
o
f
B
AR
T
+BERT
m
o
d
el
v
s
.
p
r
ev
io
u
s
AB
SA m
o
d
els
M
o
d
e
l
P
r
e
c
i
s
i
o
n
m
a
c
ro
R
e
c
a
l
l
m
a
c
r
o
F1
m
a
c
ro
B
A
R
T
+
B
E
R
T
(
p
r
o
p
o
se
d
)
0
.
6
7
0
.
7
8
0
.
7
0
B
ER
T
0
.
2
4
0
.
1
4
0
.
1
5
K
N
N
0
.
4
2
0
.
2
1
0
.
2
5
Li
n
e
a
r
S
V
M
0
.
0
5
0
.
0
6
0
.
0
5
RBF
-
S
V
M
0
.
3
1
0
.
1
1
0
.
1
3
D
e
c
i
s
i
o
n
t
r
e
e
0
.
2
8
0
.
1
5
0
.
1
7
R
a
n
d
o
m
f
o
r
e
s
t
0
.
0
8
0
.
0
6
0
.
0
5
M
LP
0
.
2
6
0
.
1
2
0
.
1
3
A
d
a
B
o
o
st
0
.
4
5
0
.
2
5
0
.
3
N
a
i
v
e
B
a
y
e
s
0
.
1
1
0
.
2
2
0
.
1
2
QDA
0
.
1
3
0
.
8
7
0
.
1
5
Ho
wev
er
,
it'
s
cr
u
cial
to
ac
k
n
o
wled
g
e
th
at
th
e
d
ata
d
is
tr
ib
u
tio
n
f
o
r
th
e
B
AR
T
+BERT
c
o
m
b
in
atio
n
d
ev
iates
f
r
o
m
n
o
r
m
ality
.
Si
m
ilar
ly
,
th
e
B
E
R
T
m
o
d
el
al
s
o
ex
h
ib
its
a
d
ev
iatio
n
f
r
o
m
n
o
r
m
al
d
is
tr
ib
u
tio
n
.
C
o
n
s
eq
u
en
tly
,
a
n
o
n
-
p
ar
am
etr
ic
W
ilco
x
o
n
s
ig
n
ed
-
r
a
n
k
test
ass
es
s
ed
th
e
s
tatis
t
ical
s
ig
n
if
i
ca
n
ce
b
etwe
en
th
e
p
r
o
p
o
s
ed
m
o
d
els
an
d
r
ec
en
t
a
p
p
r
o
ac
h
es,
s
u
ch
as
B
E
R
T
[
1
9
]
.
T
h
is
test
y
ield
ed
a
p
-
v
alu
e
o
f
0
.
0
4
1
,
in
d
icatin
g
th
e
p
-
v
alu
e
is
less
th
an
α
(
0
.
0
5
)
,
wh
ic
h
r
ev
ea
ls
th
at
th
e
o
b
s
er
v
ed
d
if
f
er
en
ce
i
n
p
er
f
o
r
m
an
ce
is
s
tatis
t
ically
s
ig
n
if
ican
t.
T
h
e
p
r
o
p
o
s
ed
m
o
d
els'
im
p
r
ess
iv
e
ca
p
ab
ilit
y
m
ak
es
it
f
ea
s
ib
le
to
ad
d
r
ess
th
e
AB
SA
ch
allen
g
e
to
ex
tr
ac
t th
e
im
p
licit ter
m
s
,
as d
ep
icted
in
T
ab
le
4
.
T
a
b
l
e
4
p
r
e
s
e
n
t
s
an
o
v
e
r
v
ie
w
o
f
s
e
v
er
a
l
f
i
n
d
i
n
g
s
o
b
t
a
in
e
d
f
r
o
m
t
h
e
r
e
s
u
l
t
s
o
f
B
A
R
T
+B
E
R
T
a
n
d
B
E
R
T
.
I
n
th
e
f
i
r
s
t
i
n
s
t
an
ce
,
B
E
R
T
i
d
en
t
i
f
i
e
s
b
o
th
"
at
t
r
a
c
t
i
o
n
"
an
d
"
i
m
ag
e
"
a
s
p
e
c
t
s
w
i
t
h
p
o
s
i
t
i
v
e
s
e
n
t
i
m
e
n
t
s
,
w
h
i
ch
d
o
e
s
n
o
t
a
l
i
g
n
w
i
th
t
h
e
e
x
p
e
c
t
ed
t
ar
g
e
t
.
B
E
R
T
f
a
i
l
s
t
o
r
e
co
g
n
i
z
e
t
h
a
t
"
m
u
s
t
v
i
s
i
t
"
i
m
p
l
i
e
s
a
r
ec
o
m
m
en
d
a
t
io
n
w
i
t
h
o
u
t
e
x
p
r
e
s
s
i
n
g
a
s
t
r
o
n
g
p
o
s
i
t
i
v
i
ty
.
Ho
w
e
v
er
,
th
e
B
A
R
T
+
B
E
R
T
m
o
d
el
s
u
c
c
e
s
s
f
u
l
ly
f
o
cu
s
e
s
o
n
i
m
p
li
c
i
t
o
p
i
n
io
n
co
n
s
i
s
t
en
t
w
i
th
t
h
e
o
v
e
r
a
l
l
co
n
t
ex
t
.
O
n
th
e
o
t
h
er
h
an
d
,
in
th
e
s
e
c
o
n
d
in
s
t
a
n
c
e,
th
e
B
E
R
T
m
o
d
e
l
i
n
ac
c
u
r
a
t
e
ly
p
r
ed
i
c
t
s
th
e
"
a
t
t
r
ac
t
i
o
n
"
a
n
d
"
im
a
g
e"
a
s
p
e
c
t
s
i
m
p
l
i
e
d
i
n
t
h
e
t
e
x
t
,
c
a
p
tu
r
in
g
o
n
l
y
th
e
ex
p
l
i
c
i
t
l
y
m
en
t
i
o
n
ed
"p
r
i
ce
"
a
s
p
e
c
t
.
M
e
an
w
h
i
l
e,
th
e
B
AR
T
+
B
E
R
T
m
o
d
e
l
s
u
c
c
e
s
s
f
u
l
ly
p
r
ed
i
c
t
s
t
h
e
i
m
p
l
i
e
d
a
s
p
e
c
t
s
d
e
s
p
i
t
e
th
e
l
o
w
r
e
a
d
ab
i
l
i
t
y
an
d
h
i
g
h
co
m
p
l
ex
i
t
y
o
f
th
e
t
e
x
t,
a
s
e
v
i
d
en
c
ed
b
y
i
t
s
h
i
g
h
F
O
G
s
co
r
e
(
2
7
.
0
2
)
an
d
lo
w
F
R
E
s
c
o
r
e
(
1
8
.
0
2
)
.
T
h
e
B
A
R
T
+
B
E
R
T
m
o
d
e
l
e
f
f
e
c
t
i
v
e
l
y
c
a
p
t
u
r
e
s
th
e
"
A
t
tr
a
c
t
io
n
"
an
d
"
I
m
a
g
e
"
a
s
p
ec
t
s
i
m
p
l
i
e
d
b
y
t
h
e
s
t
r
o
n
g
d
i
s
s
a
t
i
s
f
a
c
t
i
o
n
w
i
th
t
h
e
h
i
g
h
en
t
r
an
c
e
t
i
c
k
e
t
p
r
i
ce
s
t
o
B
o
r
o
b
u
d
u
r
.
T
h
e
r
e
s
u
l
t
s
h
ig
h
li
g
h
t
t
h
e
p
r
o
p
o
s
e
d
m
o
d
e
l
's
e
x
ce
l
l
e
n
ce
i
n
ad
d
r
e
s
s
i
n
g
s
e
n
t
im
e
n
t
a
n
a
l
y
s
i
s
a
t
t
h
e
a
s
p
ec
t
l
ev
e
l,
s
h
o
wc
a
s
i
n
g
i
t
s
ab
i
l
i
t
y
t
o
d
i
s
c
e
r
n
an
d
in
t
er
p
r
e
t
v
a
r
io
u
s
t
o
u
r
i
s
m
a
s
p
e
c
t
s
c
o
n
t
a
in
ed
w
i
t
h
in
t
ex
t
u
a
l
c
o
n
te
n
t
,
ev
en
i
n
t
h
e
c
o
n
t
ex
t
o
f
lo
w
r
e
ad
a
b
i
l
i
ty
an
d
c
o
m
p
l
ex
s
en
t
e
n
ce
s
.
T
h
e
r
e
s
u
l
t
s
h
i
g
h
l
i
g
h
t
t
h
e
p
r
o
p
o
s
ed
m
o
d
e
l's
e
x
c
e
l
l
e
n
c
e
i
n
ad
d
r
e
s
s
i
n
g
s
en
t
i
m
en
t
a
n
a
ly
s
i
s
a
t
t
h
e
a
s
p
ec
t
l
e
v
e
l
,
s
h
o
w
c
a
s
i
n
g
i
t
s
a
b
i
l
i
t
y
to
d
i
s
c
e
r
n
an
d
in
te
r
p
r
e
t
v
a
r
i
o
u
s
t
o
u
r
i
s
m
a
s
p
e
c
ts
c
o
n
t
a
in
e
d
w
i
t
h
in
t
ex
t
u
a
l
co
n
t
en
t
,
ev
en
i
n
t
h
e
c
o
n
t
ex
t
o
f
lo
w
r
e
ad
a
b
i
l
i
ty
an
d
c
o
m
p
le
x
s
en
t
e
n
c
e
s
.
T
ab
le
4
.
C
o
m
p
a
r
ativ
e
AB
SA ta
s
k
ex
am
p
le
F
R
E
F
O
G
I
n
p
u
t
t
e
x
t
P
r
e
d
i
c
t
i
o
n
s
Ta
r
g
e
t
B
A
R
T
+
B
E
R
T
B
ER
T
8
2
.
8
1
5
.
8
4
"
A
mu
st
v
i
si
t
i
n
Y
o
g
y
a
k
a
r
t
a
!
G
o
f
o
r
t
h
e
su
n
r
i
s
e
f
o
r
su
r
e
"
(
A
t
t
r
a
c
t
i
o
n
,
0
)
(
A
t
t
r
a
c
t
i
o
n
,
1
)
,
(
I
mag
e
,
1
)
(
A
t
t
r
a
c
t
i
o
n
,
0
)
1
8
.
0
2
2
7
.
0
2
"
Th
e
e
n
t
r
a
n
c
e
t
i
c
k
e
t
f
o
r
f
o
r
e
i
g
n
e
r
s
i
s
v
e
r
y
e
x
p
e
n
si
v
e
,
3
5
0
t
h
o
u
sa
n
d
,
l
o
c
a
l
5
0
t
h
o
u
sa
n
d
,
v
e
r
y
e
x
p
e
n
si
v
e
,
i
t
's
n
o
w
o
n
d
e
r
t
o
u
r
i
s
t
s
r
a
r
e
l
y
c
o
m
e
t
o
B
o
r
o
b
u
d
u
r
,
i
n
B
a
l
i
a
l
o
n
e
t
h
e
r
e
a
r
e
m
a
n
y
i
n
t
e
r
e
st
i
n
g
a
n
d
g
r
e
a
t
t
o
u
r
i
s
t
a
t
t
r
a
c
t
i
o
n
s
t
o
v
i
si
t
b
u
t
t
h
e
e
n
t
r
a
n
c
e
t
i
c
k
e
t
p
r
i
c
e
i
s
n
o
t
e
v
e
n
5
x
t
h
a
t
f
o
r
f
o
r
e
i
g
n
e
r
s
,
v
e
r
y
d
i
s
a
p
p
o
i
n
t
e
d
c
a
me
a
l
l
t
h
e
w
a
y
f
r
o
m
B
a
l
i
t
o
M
a
g
e
l
a
n
g
b
r
i
n
g
i
n
g
f
o
r
e
i
g
n
g
u
e
st
s"
(
P
r
i
c
e
,
-
1
),
(
I
mag
e
,
-
1
)
,
(
A
t
t
r
a
c
t
i
o
n
,
-
1)
(
P
r
i
c
e
,
-
1
)
(
P
r
i
c
e
,
-
1
),
(
I
mag
e
,
-
1
)
,
(
A
t
t
r
a
c
t
i
o
n
,
-
1)
3
.
2
.
P
er
f
o
r
m
a
nce
B
ART+
B
E
RT
w
it
h
t
ex
t
a
ug
m
ent
a
t
io
n
T
ab
le
5
p
r
esen
ts
a
co
m
p
ar
ati
v
e
an
aly
s
is
o
f
m
o
d
el
p
er
f
o
r
m
an
ce
ac
r
o
s
s
v
ar
io
u
s
tex
t
au
g
m
en
tatio
n
tech
n
iq
u
es.
T
h
e
r
esu
lts
im
p
ly
th
at
th
e
o
r
ig
in
al
d
ataset
with
o
u
t
au
g
m
en
tatio
n
,
d
em
o
n
s
tr
ates
g
o
o
d
p
er
f
o
r
m
an
ce
.
T
h
e
W
ilco
x
o
n
s
ig
n
ed
-
r
an
k
test
o
b
tain
ed
a
p
-
v
alu
e
o
f
0
.
0
7
4
f
o
r
wo
r
d
r
e
p
lace
m
en
t,
0
.
0
7
8
f
o
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Tr
a
n
s
fo
r
mer a
n
d
text
a
u
g
men
t
a
tio
n
fo
r
to
u
r
is
m
a
s
p
ec
t
-
b
a
s
ed
s
en
timen
t a
n
a
lysi
s
(
S
a
mu
el
S
i
tu
mea
n
g
)
4619
b
ac
k
tr
an
s
latio
n
,
an
d
0
.
0
7
4
f
o
r
r
an
d
o
m
d
eletio
n
.
No
tab
l
y
,
all
p
-
v
alu
es
ar
e
m
o
r
e
th
an
0
.
0
5
,
in
d
icatin
g
n
o
s
tatis
t
ically
s
ig
n
if
ican
t
d
if
f
er
e
n
ce
in
p
er
f
o
r
m
a
n
ce
b
etwe
en
t
h
e
B
AR
T
+BE
R
T
an
d
u
n
tr
ea
ted
m
o
d
el.
T
h
e
s
tu
d
y
f
o
u
n
d
th
at
au
g
m
e
n
tin
g
th
e
te
x
t
o
f
ten
alter
ed
o
r
ev
e
n
r
em
o
v
ed
th
e
s
p
ec
if
ic
to
u
r
is
m
c
o
n
tex
t
cr
u
cial
to
t
h
e
r
ev
iews.
T
h
is
ch
an
g
e
in
co
n
t
ex
t
r
esu
lted
in
t
h
e
m
o
d
if
icatio
n
o
f
th
e
q
u
ad
r
u
p
le
ty
p
e,
wh
ich
co
n
s
is
ts
o
f
th
e
asp
ec
t
ter
m
a
n
d
o
p
in
i
o
n
ter
m
.
C
o
n
s
eq
u
en
tly
,
th
is
i
n
tr
o
d
u
ce
d
n
o
is
e
r
ath
er
th
a
n
e
n
h
an
cin
g
th
e
d
ataset,
lead
i
n
g
to
n
o
im
p
r
o
v
e
m
en
t in
th
e
m
o
d
el's ab
ilit
y
.
T
ab
le
5
.
Per
f
o
r
m
an
ce
o
f
B
AR
T
+BERT
with
tex
t a
u
g
m
en
tatio
n
M
o
d
e
l
P
r
e
c
i
s
i
o
n
m
a
c
r
o
R
e
c
a
l
l
m
a
c
r
o
F1
m
a
c
ro
O
r
i
g
i
n
a
l
0.
67
0
.
7
8
0
.
7
0
B
a
c
k
t
r
a
n
sl
a
t
i
o
n
0
.
4
4
0
.
5
5
0
.
4
0
R
a
n
d
o
m
d
e
l
e
t
i
o
n
0
.
4
3
0
.
5
2
0
.
3
8
W
o
r
d
r
e
p
l
a
c
e
m
e
n
t
0
.
4
3
0
.
5
2
0
.
3
8
3
.
3
.
P
er
f
o
r
m
a
nce
B
ART+
B
E
RT
w
it
h
t
ex
t
prepro
ce
s
s
ing
T
ab
le
6
s
h
o
wca
s
es
a
co
m
p
ar
ativ
e
an
aly
s
is
o
f
m
o
d
el
p
er
f
o
r
m
an
ce
em
p
l
o
y
in
g
d
if
f
er
en
t
tex
t
p
r
ep
r
o
ce
s
s
in
g
tech
n
iq
u
es.
T
h
e
r
esu
lt
p
o
in
ts
to
th
e
o
r
ig
in
al
d
ataset
with
o
u
t
p
r
ep
r
o
ce
s
s
in
g
,
d
em
o
n
s
tr
atin
g
ef
f
ec
tiv
e
p
e
r
f
o
r
m
an
ce
.
T
h
e
W
ilco
x
o
n
s
ig
n
ed
-
r
an
k
test
p
r
o
d
u
ce
d
ar
e
as
f
o
llo
ws:
ca
s
e
f
o
ld
in
g
at
0
.
0
7
4
,
s
lan
g
wo
r
d
p
r
o
ce
s
s
in
g
at
0
.
0
7
4
,
d
ig
it
co
n
v
e
r
s
io
n
at
0
.
0
7
4
,
s
to
p
wo
r
d
r
em
o
v
al
at
0
.
0
7
4
,
s
p
ec
ial
ch
ar
ac
ter
r
em
o
v
al
at
0
.
0
7
8
,
lem
m
atiza
tio
n
at
0
.
0
7
8
,
a
n
d
s
tem
m
in
g
at
0
.
0
7
8
.
Sin
ce
all
p
-
v
alu
es
a
r
e
m
o
r
e
th
a
n
0
.
0
5
,
th
u
s
th
er
e
is
n
o
s
tatis
tically
s
ig
n
if
ican
t d
if
f
er
e
n
ce
in
p
e
r
f
o
r
m
an
ce
b
etwe
en
th
e
B
AR
T
+BERT
an
d
u
n
tr
ea
ted
m
o
d
el.
T
ab
le
6
.
Per
f
o
r
m
an
ce
o
f
B
AR
T
+BERT
with
tex
t p
r
ep
r
o
ce
s
s
in
g
M
o
d
e
l
P
r
e
c
i
s
i
o
n
m
a
c
r
o
R
e
c
a
l
l
m
a
c
r
o
F1
m
a
c
ro
O
r
i
g
i
n
a
l
0.
67
0
.
7
8
0
.
7
0
C
a
se
f
o
l
d
i
n
g
(
l
o
w
e
r
c
a
s
i
n
g
)
0
.
5
6
0
.
3
5
0
.
4
0
S
l
a
n
g
w
o
r
d
s
c
o
n
v
e
r
si
o
n
0
.
5
6
0
.
3
3
0
.
3
9
D
i
g
i
t
c
o
n
v
e
r
si
o
n
0
.
5
6
0
.
3
4
0
.
4
0
S
p
e
c
i
a
l
c
h
a
r
a
c
t
e
r
r
e
mo
v
a
l
0
.
5
7
0
.
3
1
0
.
3
6
S
t
o
p
w
o
r
d
r
e
m
o
v
a
l
0
.
4
7
0
.
2
3
0
.
2
7
Le
mm
a
t
i
z
a
t
i
o
n
0
.
5
7
0
.
3
1
0
.
3
6
4.
CO
NCLU
SI
O
N
Ou
r
s
tu
d
y
r
e
v
ea
ls
th
e
s
u
p
er
i
o
r
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
B
AR
T
+BERT
en
s
em
b
le
m
o
d
el
in
an
aly
zin
g
asp
ec
t
-
lev
el
s
en
tim
en
t.
T
h
e
W
ilco
x
o
n
s
ig
n
ed
-
r
an
k
test
r
ev
ea
ls
th
at
B
AR
T
+
B
E
R
T
en
s
em
b
le
m
o
d
els
d
em
o
n
s
tr
ate
s
ig
n
if
ican
tly
s
u
p
er
io
r
p
e
r
f
o
r
m
an
ce
th
a
n
n
o
n
-
e
n
s
em
b
le
m
o
d
els,
with
a
p
-
v
alu
e
o
f
0
.
0
4
1
,
u
n
d
e
r
th
e
s
ig
n
if
ican
ce
th
r
esh
o
ld
o
f
α
(
0
.
0
5
)
.
Fu
r
th
er
m
o
r
e,
ac
h
i
ev
in
g
an
F1
-
s
co
r
e
o
f
7
0
%,
th
e
p
r
o
p
o
s
ed
m
o
d
el
ex
h
ib
ited
r
o
b
u
s
t
ca
p
ab
ilit
ies
in
r
ec
o
g
n
izin
g
an
d
in
te
r
p
r
eti
n
g
to
u
r
is
m
-
r
elate
d
asp
ec
ts
w
ith
in
r
ev
iews,
ev
en
am
id
s
t
lo
w
r
ea
d
a
b
ilit
y
an
d
co
m
p
lex
s
en
ten
ce
s
.
Fu
r
t
h
er
m
o
r
e,
we
in
v
esti
g
ated
th
e
im
p
ac
t
o
f
two
s
u
p
p
lem
en
tar
y
tech
n
iq
u
es,
n
am
ely
tex
t
au
g
m
en
tatio
n
an
d
tex
t
p
r
e
p
r
o
ce
s
s
in
g
,
o
n
th
e
ca
p
ab
ilit
y
o
f
th
e
B
AR
T
+
B
E
R
T
m
o
d
el.
Su
r
p
r
is
in
g
ly
,
n
eith
er
tech
n
iq
u
e
b
r
o
u
g
h
t
im
p
r
o
v
em
e
n
ts
to
th
e
p
r
o
p
o
s
ed
m
o
d
el
p
er
f
o
r
m
an
ce
.
I
n
co
n
clu
d
in
g
o
u
r
s
tu
d
y
,
we
h
a
v
e
d
em
o
n
s
tr
ated
th
e
ef
f
ec
tiv
e
n
ess
o
f
th
e
B
AR
T
+BE
R
T
en
s
em
b
le
m
o
d
el
in
AB
SA,
ev
en
in
lo
w
r
ea
d
ab
ilit
y
an
d
co
m
p
lex
s
en
ten
ce
s
.
T
h
is
u
n
d
er
s
co
r
es
its
p
o
ten
tial
a
s
an
ef
f
icien
t
s
o
lu
tio
n
f
o
r
s
en
tim
en
t
an
aly
s
is
in
to
u
r
is
m
r
e
v
iews.
Ho
wev
er
,
f
o
r
f
u
tu
r
e
wo
r
k
e
n
d
ea
v
o
r
s
,
we
r
ec
o
m
m
en
d
ex
p
lo
r
in
g
d
o
m
ain
-
s
p
ec
if
ic
au
g
m
en
tatio
n
tech
n
i
q
u
es f
o
r
to
u
r
is
m
-
r
elate
d
d
ata
th
at
ca
n
p
r
eser
v
e
th
e
q
u
ad
r
u
p
le
ty
p
e.
Fu
r
th
e
r
m
o
r
e
,
f
u
tu
r
e
s
tu
d
ies
m
ay
also
co
n
s
id
er
au
to
m
atin
g
th
e
h
y
p
er
p
a
r
am
eter
s
co
n
f
ig
u
r
atio
n
f
o
r
th
e
B
AR
T
+BERT
en
s
em
b
le
m
o
d
el
to
m
ain
tain
its
p
er
f
o
r
m
an
ce
ed
g
e
in
AB
SA.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
e
au
th
o
r
s
wo
u
ld
lik
e
to
e
x
p
r
ess
th
eir
g
r
atitu
d
e
to
th
e
R
esear
ch
an
d
C
o
m
m
u
n
ity
Ser
v
ice
Un
it
(
L
PP
M)
o
f
I
n
s
titu
t
T
ek
n
o
lo
g
i
Del
f
o
r
-
f
u
n
d
in
g
r
esear
c
h
u
n
d
er
co
n
tr
ac
t
n
u
m
b
e
r
0
2
1
.
3
3
/I
T
Del/L
PP
M/Pen
elitia
n
/I
V/2
0
2
4
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
T
h
is
jo
u
r
n
al
u
s
es
th
e
C
o
n
tr
ib
u
to
r
R
o
les
T
ax
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
id
u
al
au
th
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
th
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
llab
o
r
atio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
4
6
1
4
-
4
6
2
2
4620
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
Sam
u
el
Sit
u
m
ea
n
g
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Sar
ah
R
o
s
d
ian
a
T
am
b
u
n
a
n
✓
✓
✓
✓
✓
✓
✓
J
ev
an
ia
✓
✓
✓
✓
✓
✓
✓
Ma
s
tawila
Feb
r
y
an
ti
Sima
n
ju
n
tak
✓
✓
✓
✓
✓
San
d
r
o
Sin
ag
a
✓
✓
✓
✓
✓
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
s
i
s
I
:
I
n
v
e
s
t
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
s
i
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
CO
NF
L
I
C
T
O
F
I
N
T
E
R
E
S
T
ST
A
T
E
M
E
NT
Au
th
o
r
s
s
tate
n
o
co
n
f
lict o
f
in
t
er
est.
DATA AV
AI
L
AB
I
L
I
T
Y
T
h
e
d
ata
th
at
s
u
p
p
o
r
t
t
h
e
f
in
d
in
g
s
o
f
th
is
s
tu
d
y
a
r
e
av
ailab
l
e
f
r
o
m
t
h
e
co
r
r
esp
o
n
d
in
g
a
u
th
o
r
,
[
SS
]
,
u
p
o
n
r
ea
s
o
n
ab
le
r
eq
u
est.
RE
F
E
R
E
NC
E
S
[
1
]
N
.
D
o
n
t
h
u
,
S
.
K
u
m
a
r
,
N
.
P
a
n
d
e
y
,
N
.
P
a
n
d
e
y
,
a
n
d
A
.
M
i
s
h
r
a
,
“
M
a
p
p
i
n
g
t
h
e
e
l
e
c
t
r
o
n
i
c
w
o
r
d
-
of
-
m
o
u
t
h
(
e
W
O
M
)
r
e
se
a
r
c
h
:
a
sy
st
e
ma
t
i
c
r
e
v
i
e
w
a
n
d
b
i
b
l
i
o
m
e
t
r
i
c
a
n
a
l
y
s
i
s,
”
J
o
u
rn
a
l
o
f
B
u
s
i
n
e
ss
Re
se
a
r
c
h
,
v
o
l
.
1
3
5
,
p
p
.
7
5
8
–
7
7
3
,
O
c
t
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
b
u
sr
e
s
.
2
0
2
1
.
0
7
.
0
1
5
.
[
2
]
H
.
S
i
m
a
n
j
u
n
t
a
k
,
D
.
T
a
r
i
g
a
n
,
I
.
S
i
b
a
r
a
n
i
,
C
.
J.
H
u
t
a
p
e
a
,
R
.
L
u
m
b
a
n
t
o
r
u
a
n
,
a
n
d
M
.
S
i
g
i
r
o
,
“
W
e
i
g
h
t
e
d
h
y
b
r
i
d
r
e
c
o
mm
e
n
d
a
t
i
o
n
sy
st
e
m
f
o
r
T
o
b
a
t
o
u
r
i
sm
b
a
s
e
d
o
n
G
o
o
g
l
e
R
e
v
i
e
w
d
a
t
a
,
”
i
n
2
0
2
2
I
E
E
E
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
I
n
f
o
rm
a
t
i
o
n
T
e
c
h
n
o
l
o
g
y
,
I
EEE,
O
c
t
.
2
0
2
2
,
p
p
.
1
–
8
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
O
S
N
I
K
O
M
5
6
5
5
1
.
2
0
2
2
.
1
0
0
3
4
9
1
1
.
[
3
]
J.
L
i
,
L.
X
u
,
L.
Ta
n
g
,
S
.
W
a
n
g
,
a
n
d
L.
L
i
,
“
B
i
g
d
a
t
a
i
n
t
o
u
r
i
sm
r
e
sea
r
c
h
:
a
l
i
t
e
r
a
t
u
r
e
r
e
v
i
e
w
,
”
T
o
u
r
i
sm
Ma
n
a
g
e
m
e
n
t
,
v
o
l
.
6
8
,
p
p
.
3
0
1
–
3
2
3
,
O
c
t
.
2
0
1
8
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
t
o
u
r
ma
n
.
2
0
1
8
.
0
3
.
0
0
9
.
[
4
]
M
.
D
.
D
e
v
i
k
a
,
C
.
S
u
n
i
t
h
a
,
a
n
d
A
.
G
a
n
e
sh
,
“
S
e
n
t
i
m
e
n
t
a
n
a
l
y
si
s
:
a
c
o
mp
a
r
a
t
i
v
e
s
t
u
d
y
o
n
d
i
f
f
e
r
e
n
t
a
p
p
r
o
a
c
h
e
s
,
”
Pr
o
c
e
d
i
a
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
v
o
l
.
8
7
,
p
p
.
4
4
–
4
9
,
2
0
1
6
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
r
o
c
s.
2
0
1
6
.
0
5
.
1
2
4
.
[
5
]
E.
M
a
r
t
i
n
i
,
F
.
T
i
sa
d
i
n
d
a
,
M
.
A
.
S
u
l
t
a
n
,
a
n
d
R
.
H
u
r
r
i
y
a
t
i
,
“
I
mp
a
c
t
o
f
e
-
W
O
M
a
n
d
W
O
M
o
n
d
e
st
i
n
a
t
i
o
n
i
ma
g
e
i
n
sh
o
p
p
i
n
g
t
o
u
r
i
sm
b
u
s
i
n
e
ss,
”
J
u
r
n
a
l
D
i
n
a
m
i
k
a
M
a
n
a
j
e
m
e
n
,
v
o
l
.
1
3
,
n
o
.
1
,
p
p
.
6
6
–
7
7
,
M
a
r
.
2
0
2
2
,
d
o
i
:
1
0
.
1
5
2
9
4
/
j
d
m
.
v
1
3
i
1
.
3
1
8
1
9
.
[
6
]
W
.
Zh
a
n
g
,
X
.
Li
,
Y
.
D
e
n
g
,
L.
B
i
n
g
,
a
n
d
W
.
La
m
,
“
A
s
u
r
v
e
y
o
n
a
s
p
e
c
t
-
b
a
se
d
se
n
t
i
me
n
t
a
n
a
l
y
si
s:
t
a
s
k
s,
me
t
h
o
d
s,
a
n
d
c
h
a
l
l
e
n
g
e
s
,
”
I
EEE
T
r
a
n
s
a
c
t
i
o
n
s
o
n
K
n
o
w
l
e
d
g
e
a
n
d
D
a
t
a
E
n
g
i
n
e
e
ri
n
g
,
v
o
l
.
3
5
,
n
o
.
1
1
,
p
p
.
1
1
0
1
9
–
1
1
0
3
8
,
N
o
v
.
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
T
K
D
E.
2
0
2
2
.
3
2
3
0
9
7
5
.
[
7
]
D
.
A
r
i
a
n
t
o
a
n
d
I
.
B
u
d
i
,
“
A
sp
e
c
t
-
b
a
s
e
d
se
n
t
i
me
n
t
a
n
a
l
y
si
s
o
n
I
n
d
o
n
e
s
i
a
’
s
t
o
u
r
i
sm
D
e
st
i
n
a
t
i
o
n
s
b
a
s
e
d
o
n
G
o
o
g
l
e
M
a
p
s
U
ser
C
o
d
e
-
mi
x
e
d
r
e
v
i
e
w
s
(
s
t
u
d
y
c
a
se
:
B
o
r
o
b
u
d
u
r
a
n
d
P
r
a
m
b
a
n
a
n
Te
m
p
l
e
s)
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
3
4
t
h
P
a
c
i
f
i
c
As
i
a
C
o
n
f
e
re
n
c
e
o
n
L
a
n
g
u
a
g
e
,
I
n
f
o
rm
a
t
i
o
n
a
n
d
C
o
m
p
u
t
a
t
i
o
n
,
H
a
n
o
i
,
V
i
e
t
n
a
m:
A
ss
o
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
s
t
i
c
s,
2
0
2
0
,
p
p
.
3
5
9
–
3
6
7
.
[
8
]
C
.
D
.
H
o
a
n
g
,
Q
.
V
.
D
i
n
h
,
a
n
d
N
.
H
.
T
r
a
n
,
“
A
s
p
e
c
t
-
c
a
t
e
g
o
r
y
-
o
p
i
n
i
o
n
-
s
e
n
t
i
m
e
n
t
e
x
t
r
a
c
t
i
o
n
u
si
n
g
g
e
n
e
r
a
t
i
v
e
t
r
a
n
sf
o
r
mer mo
d
e
l
,
”
i
n
2
0
2
2
RI
VF
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
i
n
g
a
n
d
C
o
m
m
u
n
i
c
a
t
i
o
n
T
e
c
h
n
o
l
o
g
i
e
s
,
I
EEE,
D
e
c
.
2
0
2
2
,
p
p
.
1
–
6
.
d
o
i
:
1
0
.
1
1
0
9
/
R
I
V
F
5
5
9
7
5
.
2
0
2
2
.
1
0
0
1
3
8
2
0
.
[
9
]
X
.
B
a
o
,
W
.
Z
h
o
n
g
q
i
n
g
,
X
.
J
i
a
n
g
,
R
.
X
i
a
o
,
a
n
d
S
.
Li
,
“
A
s
p
e
c
t
-
b
a
se
d
s
e
n
t
i
me
n
t
a
n
a
l
y
s
i
s
w
i
t
h
o
p
i
n
i
o
n
t
r
e
e
g
e
n
e
r
a
t
i
o
n
,
”
i
n
T
h
e
T
h
i
r
t
y
-
F
i
rst
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
i
n
t
C
o
n
f
e
re
n
c
e
o
n
Ar
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
C
a
l
i
f
o
r
n
i
a
:
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
i
n
t
C
o
n
f
e
r
e
n
c
e
s
o
n
A
r
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
O
r
g
a
n
i
z
a
t
i
o
n
,
J
u
l
.
2
0
2
2
,
p
p
.
4
0
4
4
–
4
0
5
0
,
d
o
i
:
1
0
.
2
4
9
6
3
/
i
j
c
a
i
.
2
0
2
2
/
5
6
1
.
[
1
0
]
C
.
S
h
o
r
t
e
n
a
n
d
T.
M
.
K
h
o
s
h
g
o
f
t
a
a
r
,
“
A
su
r
v
e
y
o
n
i
ma
g
e
d
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
f
o
r
d
e
e
p
l
e
a
r
n
i
n
g
,
”
J
o
u
r
n
a
l
o
f
Bi
g
D
a
t
a
,
v
o
l
.
6
,
n
o
.
1
,
p
p
.
6
0
,
D
e
c
.
2
0
1
9
,
d
o
i
:
1
0
.
1
1
8
6
/
s
4
0
5
3
7
-
0
1
9
-
0
1
9
7
-
0.
[
1
1
]
I
.
S
o
l
a
i
ma
n
e
t
a
l
.
,
“
R
e
l
e
a
se
st
r
a
t
e
g
i
e
s
a
n
d
t
h
e
s
o
c
i
a
l
i
m
p
a
c
t
s
o
f
l
a
n
g
u
a
g
e
m
o
d
e
l
s,”
a
r
Xi
v
:
1
9
0
8
.
0
9
2
0
3
,
N
o
v
.
2
0
1
9
.
[
1
2
]
S
.
Lo
n
g
p
r
e
,
Y
.
W
a
n
g
,
a
n
d
C
.
D
u
B
o
i
s
,
“
H
o
w
e
f
f
e
c
t
i
v
e
i
s
t
a
s
k
-
a
g
n
o
st
i
c
d
a
t
a
a
u
g
me
n
t
a
t
i
o
n
f
o
r
p
r
e
t
r
a
i
n
e
d
t
r
a
n
sf
o
r
mers
?
,
”
i
n
Fi
n
d
i
n
g
s
o
f
t
h
e
Ass
o
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
st
i
c
s:
EM
N
L
P
2
0
2
0
,
S
t
r
o
u
d
s
b
u
r
g
,
P
A
,
U
S
A
:
A
sso
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
Li
n
g
u
i
s
t
i
c
s,
2
0
2
0
,
p
p
.
4
4
0
1
–
4
4
1
1
,
d
o
i
:
1
0
.
1
8
6
5
3
/
v
1
/
2
0
2
0
.
f
i
n
d
i
n
g
s
-
e
m
n
l
p
.
3
9
4
.
[
1
3
]
M
.
L
e
w
i
s
e
t
a
l
.
,
“
B
A
R
T:
d
e
n
o
i
s
i
n
g
s
e
q
u
e
n
c
e
-
to
-
se
q
u
e
n
c
e
p
r
e
-
t
r
a
i
n
i
n
g
f
o
r
n
a
t
u
r
a
l
l
a
n
g
u
a
g
e
g
e
n
e
r
a
t
i
o
n
,
t
r
a
n
sl
a
t
i
o
n
,
a
n
d
c
o
m
p
r
e
h
e
n
si
o
n
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
5
8
t
h
A
n
n
u
a
l
Me
e
t
i
n
g
o
f
t
h
e
Ass
o
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
st
i
c
s
,
S
t
r
o
u
d
s
b
u
r
g
,
P
A
,
U
S
A
:
A
sso
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
Li
n
g
u
i
s
t
i
c
s,
2
0
2
0
,
p
p
.
7
8
7
1
–
7
8
8
0
,
d
o
i
:
1
0
.
1
8
6
5
3
/
v
1
/
2
0
2
0
.
a
c
l
-
m
a
i
n
.
7
0
3
.
[
1
4
]
G
.
W
a
n
g
,
J
.
H
a
o
,
J.
M
a
,
a
n
d
H
.
J
i
a
n
g
,
“
A
c
o
mp
a
r
a
t
i
v
e
a
ss
e
ssm
e
n
t
o
f
e
n
se
mb
l
e
l
e
a
r
n
i
n
g
f
o
r
c
r
e
d
i
t
s
c
o
r
i
n
g
,
”
Ex
p
e
r
t
S
y
st
e
m
s
w
i
t
h
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
3
8
,
n
o
.
1
,
p
p
.
2
2
3
–
2
3
0
,
Ja
n
.
2
0
1
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
sw
a
.
2
0
1
0
.
0
6
.
0
4
8
.
[
1
5
]
D
.
O
p
i
t
z
a
n
d
R
.
M
a
c
l
i
n
,
“
P
o
p
u
l
a
r
e
n
semb
l
e
m
e
t
h
o
d
s
:
a
n
e
mp
i
r
i
c
a
l
st
u
d
y
,
”
J
o
u
r
n
a
l
o
f
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
R
e
se
a
rc
h
,
v
o
l
.
1
1
,
n
o
.
1
,
p
p
.
1
6
9
–
1
9
8
,
1
9
9
9
.
[
1
6
]
J.
M
.
M
o
r
e
i
r
a
,
A
.
C
.
P
.
L
.
F
.
d
e
C
a
r
v
a
l
h
o
,
a
n
d
T.
H
o
r
v
á
t
h
,
“
D
a
t
a
q
u
a
l
i
t
y
a
n
d
p
r
e
p
r
o
c
e
ssi
n
g
,
”
i
n
A
G
e
n
e
r
a
l
I
n
t
r
o
d
u
c
t
i
o
n
t
o
D
a
t
a
An
a
l
y
t
i
c
s
,
W
i
l
e
y
,
2
0
1
8
,
p
p
.
7
1
–
97
,
d
o
i
:
1
0
.
1
0
0
2
/
9
7
8
1
1
1
9
2
9
6
2
9
4
.
c
h
4
.
[
1
7
]
N
.
I
n
d
u
r
k
h
y
a
a
n
d
F
.
J.
D
a
m
e
r
a
u
,
H
a
n
d
b
o
o
k
o
f
n
a
t
u
ra
l
l
a
n
g
u
a
g
e
p
ro
c
e
ss
i
n
g
,
2
n
d
e
d
.
N
e
w
Y
o
r
k
:
C
h
a
p
m
a
n
a
n
d
H
a
l
l
/
C
R
C
,
2
0
1
0
,
d
o
i
:
1
0
.
1
2
0
1
/
9
7
8
1
4
2
0
0
8
5
9
3
8
.
[
1
8
]
M
.
J.
D
e
n
n
y
a
n
d
A
.
S
p
i
r
l
i
n
g
,
“
Te
x
t
p
r
e
p
r
o
c
e
ssi
n
g
f
o
r
u
n
su
p
e
r
v
i
s
e
d
l
e
a
r
n
i
n
g
:
w
h
y
i
t
m
a
t
t
e
r
s,
w
h
e
n
i
t
mi
s
l
e
a
d
s
,
a
n
d
w
h
a
t
t
o
d
o
a
b
o
u
t
i
t
,
”
Po
l
i
t
i
c
a
l
A
n
a
l
y
s
i
s
,
v
o
l
.
2
6
,
n
o
.
2
,
p
p
.
1
6
8
–
1
8
9
,
A
p
r
.
2
0
1
8
,
d
o
i
:
1
0
.
1
0
1
7
/
p
a
n
.
2
0
1
7
.
4
4
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
Tr
a
n
s
fo
r
mer a
n
d
text
a
u
g
men
t
a
tio
n
fo
r
to
u
r
is
m
a
s
p
ec
t
-
b
a
s
ed
s
en
timen
t a
n
a
lysi
s
(
S
a
mu
el
S
i
tu
mea
n
g
)
4621
[
1
9
]
J.
D
e
v
l
i
n
,
M
.
-
W
.
C
h
a
n
g
,
K
.
Le
e
,
a
n
d
K
.
To
u
t
a
n
o
v
a
,
“
B
E
R
T
:
P
r
e
-
t
r
a
i
n
i
n
g
o
f
d
e
e
p
b
i
d
i
r
e
c
t
i
o
n
a
l
t
r
a
n
sf
o
r
m
e
r
s
f
o
r
l
a
n
g
u
a
g
e
u
n
d
e
r
s
t
a
n
d
i
n
g
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
N
o
rt
h
Am
e
ri
c
a
n
C
h
a
p
t
e
r
o
f
t
h
e
Ass
o
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
st
i
c
s
,
S
t
r
o
u
d
s
b
u
r
g
,
P
A
,
U
S
A
:
A
sso
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
Li
n
g
u
i
s
t
i
c
s,
2
0
1
9
,
p
p
.
4
1
7
1
–
4
1
8
6
,
d
o
i
:
1
0
.
1
8
6
5
3
/
v
1
/
N
1
9
-
1
4
2
3
.
[
2
0
]
E.
C
.
G
.
-
M
e
r
c
h
a
n
,
R
.
G
.
-
B
r
i
z
u
e
l
a
,
a
n
d
S
.
G
.
-
C
a
r
v
a
j
a
l
,
“
C
o
m
p
a
r
i
n
g
B
E
R
T
a
g
a
i
n
st
t
r
a
d
i
t
i
o
n
a
l
ma
c
h
i
n
e
l
e
a
r
n
i
n
g
mo
d
e
l
s
i
n
t
e
x
t
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
J
o
u
r
n
a
l
o
f
C
o
m
p
u
t
a
t
i
o
n
a
l
a
n
d
C
o
g
n
i
t
i
v
e
En
g
i
n
e
e
r
i
n
g
,
v
o
l
.
2
,
n
o
.
4
,
p
p
.
3
5
2
–
3
5
6
,
A
p
r
.
2
0
2
3
,
d
o
i
:
1
0
.
4
7
8
5
2
/
b
o
n
v
i
e
w
JC
C
E
3
2
0
2
8
3
8
.
[
2
1
]
M
.
A
r
k
h
i
p
o
v
,
M
.
Tr
o
f
i
mo
v
a
,
Y
.
K
u
r
a
t
o
v
,
a
n
d
A
.
S
o
r
o
k
i
n
,
“
T
u
n
i
n
g
m
u
l
t
i
l
i
n
g
u
a
l
t
r
a
n
sf
o
r
m
e
r
s
f
o
r
l
a
n
g
u
a
g
e
-
s
p
e
c
i
f
i
c
n
a
me
d
e
n
t
i
t
y
r
e
c
o
g
n
i
t
i
o
n
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
7
t
h
Wo
r
k
sh
o
p
o
n
Ba
l
t
o
-
S
l
a
v
i
c
N
a
t
u
ra
l
L
a
n
g
u
a
g
e
Pro
c
e
ss
i
n
g
,
S
t
r
o
u
d
s
b
u
r
g
,
P
A
,
U
S
A
:
A
sso
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
st
i
c
s
,
2
0
1
9
,
p
p
.
8
9
–
93
,
d
o
i
:
1
0
.
1
8
6
5
3
/
v
1
/
W
1
9
-
3
7
1
2
.
[
2
2
]
V
.
G
o
y
a
l
,
R
.
S
i
n
g
h
,
M
.
D
h
a
w
l
e
y
,
A
.
K
u
m
a
r
,
a
n
d
S
.
S
h
a
r
ma,
“
A
e
r
i
a
l
o
b
j
e
c
t
d
e
t
e
c
t
i
o
n
u
s
i
n
g
d
e
e
p
l
e
a
r
n
i
n
g
:
a
r
e
v
i
e
w
,
”
i
n
C
o
m
p
u
t
a
t
i
o
n
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
S
p
r
i
n
g
e
r
,
S
i
n
g
a
p
o
r
e
,
2
0
2
3
,
p
p
.
8
1
–
92
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
9
8
1
-
19
-
7
3
4
6
-
8
_
8
.
[
2
3
]
G.
-
M
.
P
a
r
k
,
S
.
-
E.
H
o
n
g
,
a
n
d
S
.
-
B
.
P
a
r
k
,
“
P
o
st
-
t
r
a
i
n
i
n
g
w
i
t
h
i
n
t
e
r
r
o
g
a
t
i
v
e
s
e
n
t
e
n
c
e
s
f
o
r
e
n
h
a
n
c
i
n
g
B
A
R
T
-
b
a
s
e
d
K
o
r
e
a
n
q
u
e
st
i
o
n
g
e
n
e
r
a
t
o
r
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
2
n
d
C
o
n
f
e
re
n
c
e
o
f
t
h
e
As
i
a
-
P
a
c
i
f
i
c
C
h
a
p
t
e
r
o
f
t
h
e
Ass
o
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
st
i
c
s
a
n
d
t
h
e
1
2
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
i
n
t
C
o
n
f
e
re
n
c
e
o
n
N
a
t
u
r
a
l
L
a
n
g
u
a
g
e
Pr
o
c
e
ss
i
n
g
,
A
sso
c
i
a
t
i
o
n
f
o
r
C
o
mp
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
st
i
c
s
,
2
0
2
2
,
p
p
.
2
0
2
–
2
0
9
.
[
2
4
]
Z.
Li
n
e
t
a
l
.
,
“
Te
x
t
g
e
n
e
r
a
t
i
o
n
w
i
t
h
d
i
f
f
u
s
i
o
n
l
a
n
g
u
a
g
e
m
o
d
e
l
s
:
a
p
r
e
-
t
r
a
i
n
i
n
g
a
p
p
r
o
a
c
h
w
i
t
h
c
o
n
t
i
n
u
o
u
s
p
a
r
a
g
r
a
p
h
d
e
n
o
i
se,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
4
0
t
h
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
M
a
c
h
i
n
e
L
e
a
rn
i
n
g
,
A
C
M
D
i
g
i
t
a
l
Li
b
r
a
r
y
,
2
0
2
3
,
p
p
.
2
1
0
5
1
–
2
1
0
6
4
.
[
2
5
]
B
.
L
i
,
Y
.
H
o
u
,
a
n
d
W
.
C
h
e
,
“
D
a
t
a
a
u
g
men
t
a
t
i
o
n
a
p
p
r
o
a
c
h
e
s i
n
n
a
t
u
r
a
l
l
a
n
g
u
a
g
e
p
r
o
c
e
ssi
n
g
:
a
s
u
r
v
e
y
,
”
AI
O
p
e
n
,
v
o
l
.
3
,
p
p
.
7
1
–
9
0
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
a
i
o
p
e
n
.
2
0
2
2
.
0
3
.
0
0
1
.
[
2
6
]
J.
C
h
e
n
,
D
.
Ta
m
,
C
.
R
a
f
f
e
l
,
M
.
B
a
n
sal
,
a
n
d
D
.
Y
a
n
g
,
“
A
n
e
m
p
i
r
i
c
a
l
s
u
r
v
e
y
o
f
d
a
t
a
a
u
g
m
e
n
t
a
t
i
o
n
f
o
r
l
i
m
i
t
e
d
d
a
t
a
l
e
a
r
n
i
n
g
i
n
N
LP,”
T
r
a
n
s
a
c
t
i
o
n
s
o
f
t
h
e
Ass
o
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
st
i
c
s
,
v
o
l
.
1
1
,
p
p
.
1
9
1
–
2
1
1
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
6
2
/
t
a
c
l
_
a
_
0
0
5
4
2
.
[
2
7
]
M
.
Q
u
e
u
d
o
t
,
L.
M
a
r
c
e
a
u
,
R
.
M
.
B
e
l
b
a
h
a
r
,
E.
C
h
a
r
t
o
n
,
a
n
d
M
.
-
J.
M
e
u
r
s,
“
D
a
t
a
s
e
t
a
u
g
me
n
t
a
t
i
o
n
u
si
n
g
b
a
c
k
-
t
r
a
n
sl
a
t
i
o
n
t
o
i
mp
r
o
v
e
e
a
r
l
y
st
a
g
e
d
i
a
l
o
g
sy
st
e
ms,”
i
n
Pr
o
c
e
e
d
i
n
g
s o
f
t
h
e
C
a
n
a
d
i
a
n
C
o
n
f
e
re
n
c
e
o
n
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
C
a
n
a
d
i
a
n
A
r
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
A
sso
c
i
a
t
i
o
n
,
M
a
y
2
0
2
2
,
d
o
i
:
1
0
.
2
1
4
2
8
/
5
9
4
7
5
7
d
b
.
3
9
8
6
f
a
3
7
.
[
2
8
]
S
.
X
i
a
o
a
n
d
J
.
W
.
O
l
l
e
r
,
“
C
a
n
r
e
l
a
t
i
v
e
l
y
p
e
r
f
e
c
t
t
r
a
n
s
l
a
t
i
o
n
b
e
t
w
e
e
n
E
n
g
l
i
sh
a
n
d
C
h
i
n
e
s
e
b
e
a
c
h
i
e
v
e
d
?
,
”
L
a
n
g
u
a
g
e
T
e
s
t
i
n
g
,
v
o
l
.
1
1
,
n
o
.
3
,
p
p
.
2
6
7
–
2
8
9
,
N
o
v
.
1
9
9
4
,
d
o
i
:
1
0
.
1
1
7
7
/
0
2
6
5
5
3
2
2
9
4
0
1
1
0
0
3
0
3
.
[
2
9
]
S
.
Zh
a
o
a
n
d
D
.
Z
h
a
n
g
,
“
T
h
e
t
o
t
a
l
i
t
y
o
f
C
h
i
n
e
s
e
c
h
a
r
a
c
t
e
r
s
-
a
d
i
g
i
t
a
l
p
e
r
sp
e
c
t
i
v
e
,
”
J
o
u
rn
a
l
o
f
C
h
i
n
e
se
L
a
n
g
u
a
g
e
a
n
d
C
o
m
p
u
t
i
n
g
,
v
o
l
.
1
7
,
n
o
.
2
,
p
p
.
1
0
7
–
1
2
5
,
2
0
0
7
.
[
3
0
]
M
.
M
u
n
i
k
a
r
,
S
.
S
h
a
k
y
a
,
a
n
d
A
.
S
h
r
e
s
t
h
a
,
“
F
i
n
e
-
g
r
a
i
n
e
d
se
n
t
i
me
n
t
c
l
a
ss
i
f
i
c
a
t
i
o
n
u
si
n
g
B
E
R
T,
”
i
n
2
0
1
9
Ar
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
f
o
r
T
ra
n
s
f
o
rm
i
n
g
Bu
si
n
e
ss
a
n
d
S
o
c
i
e
t
y
,
I
EEE,
N
o
v
.
2
0
1
9
,
p
p
.
1
–
5
,
d
o
i
:
1
0
.
1
1
0
9
/
A
I
TB
4
8
5
1
5
.
2
0
1
9
.
8
9
4
7
4
3
5
.
[
3
1
]
S
.
S
y
me
o
n
i
d
i
s
,
D
.
Ef
f
r
o
s
y
n
i
d
i
s
,
a
n
d
A
.
A
r
a
mp
a
t
z
i
s,
“
A
c
o
mp
a
r
a
t
i
v
e
e
v
a
l
u
a
t
i
o
n
o
f
p
r
e
-
p
r
o
c
e
ssi
n
g
t
e
c
h
n
i
q
u
e
s
a
n
d
t
h
e
i
r
i
n
t
e
r
a
c
t
i
o
n
s
f
o
r
Tw
i
t
t
e
r
s
e
n
t
i
m
e
n
t
a
n
a
l
y
si
s
,
”
Ex
p
e
r
t
S
y
s
t
e
m
s
w
i
t
h
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
1
1
0
,
p
p
.
2
9
8
–
3
1
0
,
N
o
v
.
2
0
1
8
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
sw
a
.
2
0
1
8
.
0
6
.
0
2
2
.
[
3
2
]
H
.
F
a
n
g
,
G
.
X
u
,
Y
.
L
o
n
g
,
a
n
d
W
.
Ta
n
g
,
“
A
n
Ef
f
e
c
t
i
v
e
EL
EC
T
R
A
-
b
a
s
e
d
p
i
p
e
l
i
n
e
f
o
r
s
e
n
t
i
me
n
t
a
n
a
l
y
s
i
s
o
f
t
o
u
r
i
s
t
a
t
t
r
a
c
t
i
o
n
r
e
v
i
e
w
s,”
A
p
p
l
i
e
d
S
c
i
e
n
c
e
s
,
v
o
l
.
1
2
,
n
o
.
2
1
,
O
c
t
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
a
p
p
1
2
2
1
1
0
8
8
1
.
[
3
3
]
M
.
M
.
S
a
mi
a
,
A
.
R
a
j
e
e
,
M
.
R
.
H
a
s
a
n
,
M
.
O
.
F
a
r
u
q
,
a
n
d
P
.
C
.
P
a
u
l
,
“
A
s
p
e
c
t
-
b
a
se
d
s
e
n
t
i
m
e
n
t
a
n
a
l
y
si
s
f
o
r
B
e
n
g
a
l
i
t
e
x
t
u
si
n
g
b
i
d
i
r
e
c
t
i
o
n
a
l
e
n
c
o
d
e
r
r
e
p
r
e
se
n
t
a
t
i
o
n
s
f
r
o
m
t
r
a
n
sf
o
r
mers
(
B
ER
T)
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
Ad
v
a
n
c
e
d
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
1
3
,
n
o
.
1
2
,
2
0
2
2
,
d
o
i
:
1
0
.
1
4
5
6
9
/
I
JA
C
S
A
.
2
0
2
2
.
0
1
3
1
2
1
1
2
.
[
3
4
]
J.
Ta
o
,
L.
Zh
o
u
,
a
n
d
C
.
F
e
e
n
e
y
,
“
I
u
n
d
e
r
st
a
n
d
w
h
a
t
y
o
u
a
r
e
s
a
y
i
n
g
:
l
e
v
e
r
a
g
i
n
g
d
e
e
p
l
e
a
r
n
i
n
g
t
e
c
h
n
i
q
u
e
s
f
o
r
a
s
p
e
c
t
b
a
s
e
d
sen
t
i
m
e
n
t
a
n
a
l
y
s
i
s,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
5
2
n
d
H
a
w
a
i
i
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
S
y
st
e
m
S
c
i
e
n
c
e
s
,
G
r
a
n
d
W
a
i
l
e
a
,
H
a
w
a
i
i
:
H
I
C
S
S
,
2
0
1
9
,
p
p
.
4
7
0
-
479
.
[
3
5
]
A
.
M
u
r
t
a
d
h
a
,
B
.
W
e
n
,
S
.
P
a
n
,
J.
S
u
,
L.
A
o
,
a
n
d
Y
.
Li
u
,
“
B
ER
T
-
A
S
C
:
a
u
x
i
l
i
a
r
y
-
se
n
t
e
n
c
e
c
o
n
s
t
r
u
c
t
i
o
n
f
o
r
i
m
p
l
i
c
i
t
a
sp
e
c
t
l
e
a
r
n
i
n
g
i
n
se
n
t
i
me
n
t
a
n
a
l
y
s
i
s,
”
E
x
p
e
rt
S
y
st
e
m
s w
i
t
h
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
2
5
8
,
D
e
c
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
sw
a
.
2
0
2
4
.
1
2
5
1
9
5
.
[
3
6
]
C
.
S
u
n
,
L.
H
u
a
n
g
,
a
n
d
X
.
Q
i
u
,
“
U
t
i
l
i
z
i
n
g
B
E
R
T
f
o
r
a
s
p
e
c
t
-
b
a
se
d
s
e
n
t
i
men
t
a
n
a
l
y
si
s
v
i
a
c
o
n
st
r
u
c
t
i
n
g
a
u
x
i
l
i
a
r
y
s
e
n
t
e
n
c
e
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
2
0
1
9
C
o
n
f
e
re
n
c
e
o
f
t
h
e
N
o
r
t
h
Am
e
r
i
c
a
n
C
h
a
p
t
e
r
o
f
t
h
e
Ass
o
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
st
i
c
s:
H
u
m
a
n
L
a
n
g
u
a
g
e
T
e
c
h
n
o
l
o
g
i
e
s
,
S
t
r
o
u
d
sb
u
r
g
,
P
A
,
U
S
A
:
A
sso
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
s
t
i
c
s,
2
0
1
9
,
p
p
.
3
8
0
–
3
8
5
.
d
o
i
:
1
0
.
1
8
6
5
3
/
v
1
/
N
1
9
-
1
0
3
5
.
[
3
7
]
A
.
D
a
s
a
n
d
W
.
E.
Zh
a
n
g
,
“
A
B
S
A
-
B
e
n
c
h
:
t
o
w
a
r
d
s
t
h
e
u
n
i
f
i
e
d
e
v
a
l
u
a
t
i
o
n
o
f
a
s
p
e
c
t
-
b
a
s
e
d
s
e
n
t
i
m
e
n
t
a
n
a
l
y
si
s
r
e
se
a
r
c
h
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
1
8
t
h
A
n
n
u
a
l
W
o
rks
h
o
p
o
f
t
h
e
A
u
st
r
a
l
a
si
a
n
L
a
n
g
u
a
g
e
T
e
c
h
n
o
l
o
g
y
Ass
o
c
i
a
t
i
o
n
,
A
u
s
t
r
a
l
a
si
a
n
L
a
n
g
u
a
g
e
Te
c
h
n
o
l
o
g
y
A
ss
o
c
i
a
t
i
o
n
,
2
0
2
0
,
p
p
.
6
5
–
7
1
.
[
3
8
]
R
.
F
l
e
sc
h
,
“
‘
S
i
m
p
l
i
f
i
c
a
t
i
o
n
o
f
F
l
e
s
c
h
r
e
a
d
i
n
g
e
a
s
e
f
o
r
mu
l
a
’
:
r
e
p
l
y
.
,
”
J
o
u
r
n
a
l
o
f
Ap
p
l
i
e
d
Psy
c
h
o
l
o
g
y
,
v
o
l
.
3
6
,
n
o
.
1
,
p
p
.
5
4
–
5
5
,
1
9
5
2
,
d
o
i
:
1
0
.
1
0
3
7
/
h
0
0
5
1
9
6
5
.
[
3
9
]
R
.
G
u
n
n
i
n
g
,
“
T
h
e
f
o
g
i
n
d
e
x
a
f
t
e
r
t
w
e
n
t
y
y
e
a
r
s,”
J
o
u
r
n
a
l
o
f
B
u
si
n
e
s
s
C
o
m
m
u
n
i
c
a
t
i
o
n
,
v
o
l
.
6
,
n
o
.
2
,
p
p
.
3
–
1
3
,
J
a
n
.
1
9
6
9
,
d
o
i
:
1
0
.
1
1
7
7
/
0
0
2
1
9
4
3
6
6
9
0
0
6
0
0
2
0
2
.
[
4
0
]
E.
F
i
x
a
n
d
J
.
L.
H
o
d
g
e
s,
“
D
i
scri
mi
n
a
t
o
r
y
a
n
a
l
y
s
i
s.
n
o
n
p
a
r
a
m
e
t
r
i
c
d
i
scri
mi
n
a
t
i
o
n
:
c
o
n
si
s
t
e
n
c
y
p
r
o
p
e
r
t
i
e
s,
”
I
n
t
e
rn
a
t
i
o
n
a
l
S
t
a
t
i
st
i
c
a
l
Re
v
i
e
w
/
R
e
v
u
e
I
n
t
e
r
n
a
t
i
o
n
a
l
e
d
e
S
t
a
t
i
s
t
i
q
u
e
,
v
o
l
.
5
7
,
n
o
.
3
,
p
p
.
2
3
8
–
2
4
7
,
D
e
c
.
1
9
8
9
,
d
o
i
:
1
0
.
2
3
0
7
/
1
4
0
3
7
9
7
.
[
4
1
]
B
.
E.
B
o
s
e
r
,
I
.
M
.
G
u
y
o
n
,
a
n
d
V
.
N
.
V
a
p
n
i
k
,
“
A
t
r
a
i
n
i
n
g
a
l
g
o
r
i
t
h
m
f
o
r
o
p
t
i
ma
l
marg
i
n
c
l
a
ssi
f
i
e
r
s,”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
f
i
f
t
h
a
n
n
u
a
l
w
o
rks
h
o
p
o
n
C
o
m
p
u
t
a
t
i
o
n
a
l
l
e
a
r
n
i
n
g
t
h
e
o
r
y
,
N
e
w
Y
o
r
k
,
U
n
i
t
e
d
S
t
a
t
e
s
:
A
C
M
,
Ju
l
.
1
9
9
2
,
p
p
.
1
4
4
–
1
5
2
.
d
o
i
:
1
0
.
1
1
4
5
/
1
3
0
3
8
5
.
1
3
0
4
0
1
.
[
4
2
]
B.
-
C
.
K
u
o
,
H
.
-
H
.
H
o
,
C
.
-
H
.
Li
,
C
.
-
C
.
H
u
n
g
,
a
n
d
J.
-
S
.
Ta
u
r
,
“
A
k
e
r
n
e
l
-
b
a
s
e
d
f
e
a
t
u
r
e
se
l
e
c
t
i
o
n
m
e
t
h
o
d
f
o
r
S
V
M
w
i
t
h
R
B
F
k
e
r
n
e
l
f
o
r
h
y
p
e
r
s
p
e
c
t
r
a
l
i
m
a
g
e
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
I
E
EE
J
o
u
r
n
a
l
o
f
S
e
l
e
c
t
e
d
T
o
p
i
c
s
i
n
A
p
p
l
i
e
d
Ea
r
t
h
O
b
ser
v
a
t
i
o
n
s
a
n
d
Re
m
o
t
e
S
e
n
si
n
g
,
v
o
l
.
7
,
n
o
.
1
,
p
p
.
3
1
7
–
3
2
6
,
J
a
n
.
2
0
1
4
,
d
o
i
:
1
0
.
1
1
0
9
/
JS
TA
R
S
.
2
0
1
3
.
2
2
6
2
9
2
6
.
[
4
3
]
J
.
R
.
Q
u
i
n
l
a
n
,
“
I
n
d
u
c
t
i
o
n
o
f
d
e
c
i
s
i
o
n
t
r
e
e
s
,
”
M
a
c
h
i
n
e
L
e
a
r
n
i
n
g
,
v
o
l
.
1
,
n
o
.
1
,
p
p
.
8
1
–
1
0
6
,
M
a
r
.
1
9
8
6
,
d
o
i
:
1
0
.
1
0
2
3
/
A
:
1
0
2
2
6
4
3
2
0
4
8
7
7
.
[
4
4
]
L.
B
r
e
i
m
a
n
,
“
R
a
n
d
o
m fo
r
e
st
s,
”
Ma
c
h
i
n
e
L
e
a
r
n
i
n
g
,
v
o
l
.
4
5
,
n
o
.
1
,
p
p
.
5
–
3
2
,
O
c
t
.
2
0
0
1
,
d
o
i
:
1
0
.
1
0
2
3
/
A
:
1
0
1
0
9
3
3
4
0
4
3
2
4
.
[
4
5
]
M.
-
C
.
P
o
p
e
s
c
u
,
V
.
E
.
B
a
l
a
s,
L
.
P
.
-
P
o
p
e
sc
u
,
a
n
d
N
.
M
a
s
t
o
r
a
k
i
s,
“
M
u
l
t
i
l
a
y
e
r
p
e
r
c
e
p
t
r
o
n
a
n
d
n
e
u
r
a
l
n
e
t
w
o
r
k
s,
”
W
S
EA
S
T
ra
n
s
a
c
t
i
o
n
s
o
n
C
i
rc
u
i
t
s
a
n
d
S
y
st
e
m
s
,
v
o
l
.
8
,
n
o
.
7
,
p
p
.
5
7
9
–
5
8
8
,
2
0
0
9
.
[
4
6
]
Y
.
F
r
e
u
n
d
a
n
d
R
.
E.
S
c
h
a
p
i
r
e
,
“
A
s
h
o
r
t
i
n
t
r
o
d
u
c
t
i
o
n
t
o
b
o
o
s
t
i
n
g
,
”
J
o
u
r
n
a
l
o
f
J
a
p
a
n
e
s
e
S
o
c
i
e
t
y
f
o
r
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
1
4
,
n
o
.
5
,
p
p
.
7
7
1
–
7
8
0
,
1
9
9
9
.
[
4
7
]
G
.
I
.
W
e
b
b
,
“
N
a
ï
v
e
B
a
y
e
s,
”
i
n
En
c
y
c
l
o
p
e
d
i
a
o
f
M
a
c
h
i
n
e
L
e
a
r
n
i
n
g
,
B
o
st
o
n
,
M
A
:
S
p
r
i
n
g
e
r
U
S
,
2
0
1
1
,
p
p
.
7
1
3
–
7
1
4
.
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
0
-
387
-
3
0
1
6
4
-
8
_
5
7
6
.
[
4
8
]
Y
.
W
u
,
Y
.
Q
i
n
,
a
n
d
M
.
Z
h
u
,
“
Q
u
a
d
r
a
t
i
c
D
i
scri
m
i
n
a
n
t
a
n
a
l
y
si
s
f
o
r
h
i
g
h
-
d
i
me
n
si
o
n
a
l
d
a
t
a,”
S
t
a
t
i
s
t
i
c
a
S
i
n
i
c
a
,
v
o
l
.
2
9
,
n
o
.
2
,
p
p
.
9
3
9
–
9
6
0
,
2
0
1
9
,
d
o
i
:
1
0
.
5
7
0
5
/
ss.
2
0
2
0
1
6
.
0
0
3
4
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
4
,
No
.
6
,
Dec
em
b
er
2
0
2
5
:
4
6
1
4
-
4
6
2
2
4622
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
S
a
m
u
e
l
S
itu
m
e
a
n
g
re
c
e
iv
e
d
h
is
b
a
c
h
e
lo
r'
s
d
e
g
re
e
in
t
h
e
In
fo
rm
a
ti
o
n
Tec
h
n
o
l
o
g
y
S
tu
d
y
P
ro
g
ra
m
,
F
a
c
u
lt
y
o
f
C
o
m
p
u
ter
S
c
ien
c
e
a
n
d
In
f
o
rm
a
ti
o
n
Tec
h
n
o
l
o
g
y
,
Un
i
v
e
rsitas
S
u
m
a
tera
Uta
ra
,
In
d
o
n
e
sia
,
in
2
0
1
4
.
He
re
c
e
iv
e
d
h
is
m
a
ste
r'
s
d
e
g
r
e
e
fro
m
th
e
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter
S
c
ien
c
e
a
n
d
I
n
fo
rm
a
t
io
n
E
n
g
i
n
e
e
rin
g
,
Co
ll
e
g
e
o
f
En
g
in
e
e
rin
g
,
Na
ti
o
n
a
l
Ch
u
n
g
Ch
e
n
g
Un
i
v
e
rsity
,
Taiwa
n
,
i
n
2
0
1
7
.
He
c
u
rre
n
tl
y
lec
tu
re
s
i
n
th
e
In
f
o
rm
a
ti
o
n
S
y
ste
m
S
t
u
d
y
P
ro
g
ra
m
,
F
a
c
u
lt
y
o
f
In
f
o
rm
a
ti
c
s
a
n
d
El
e
c
tri
c
a
l
En
g
i
n
e
e
rin
g
,
I
n
stit
u
t
Tek
n
o
lo
g
i
De
l.
His
m
a
in
re
se
a
rc
h
in
tere
st
is
AI
-
d
ri
v
e
n
so
lu
ti
o
n
s
f
o
r
e
d
u
c
a
ti
o
n
a
n
d
c
u
lt
u
r
a
l
h
e
r
it
a
g
e
,
w
h
ich
in
ters
e
c
t
a
re
a
s
o
f
d
a
ta
sc
ien
c
e
,
n
a
t
u
ra
l
lan
g
u
a
g
e
p
ro
c
e
ss
in
g
,
a
n
d
c
o
m
p
u
ter
v
isio
n
.
He
c
a
n
b
e
c
o
n
tac
te
d
a
t
e
m
a
il
:
sa
m
u
e
l.
situ
m
e
a
n
g
@d
e
l.
a
c
.
id
.
S
a
r
a
h
Ro
sdi
a
n
a
T
a
m
b
u
n
a
n
re
c
e
iv
e
d
h
e
r
b
a
c
h
e
lo
r'
s
d
e
g
re
e
i
n
t
h
e
I
n
fo
rm
a
ti
o
n
S
y
ste
m
S
tu
d
y
P
r
o
g
ra
m
,
F
a
c
u
lt
y
o
f
In
fo
rm
a
ti
c
s
a
n
d
El
e
c
tri
c
a
l
En
g
in
e
e
rin
g
,
In
stit
u
t
Tek
n
o
l
o
g
i
De
l,
In
d
o
n
e
sia
,
in
2
0
1
9
.
S
h
e
re
c
e
iv
e
d
h
e
r
m
a
ste
r'
s
d
e
g
re
e
fro
m
th
e
Co
m
p
u
ter
S
c
ie
n
c
e
s
M
a
ste
r
P
ro
g
ra
m
,
F
a
c
u
lt
y
o
f
M
a
t
h
e
m
a
ti
c
s
a
n
d
Na
tu
ra
l
S
c
ien
c
e
s,
Un
i
v
e
rsitas
G
a
d
jah
M
a
d
a
,
I
n
d
o
n
e
sia
,
in
2
0
2
3
.
S
h
e
c
u
rre
n
tl
y
lec
tu
re
s
in
t
h
e
In
f
o
rm
a
ti
o
n
S
y
ste
m
S
t
u
d
y
P
r
o
g
ra
m
,
F
a
c
u
lt
y
o
f
In
fo
rm
a
ti
c
s
a
n
d
El
e
c
tri
c
a
l
E
n
g
i
n
e
e
rin
g
,
I
n
stit
u
t
Te
k
n
o
lo
g
i
De
l.
He
r
re
se
a
rc
h
fo
c
u
se
s
o
n
so
ftwa
re
e
n
g
in
e
e
rin
g
a
n
d
d
a
ta
se
c
u
rit
y
.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
sa
ra
h
.
tam
b
u
n
a
n
@
d
e
l.
a
c
.
id
.
J
e
v
a
n
ia
re
c
e
iv
e
d
h
e
r
b
a
c
h
e
lo
r'
s
d
e
g
re
e
i
n
th
e
In
f
o
rm
a
ti
o
n
S
y
ste
m
S
tu
d
y
P
r
o
g
ra
m
,
F
a
c
u
lt
y
o
f
I
n
fo
rm
a
ti
c
s
a
n
d
El
e
c
tr
ica
l
En
g
i
n
e
e
rin
g
,
In
st
it
u
t
Tek
n
o
l
o
g
i
De
l,
In
d
o
n
e
sia
,
in
2
0
2
4
.
He
r
c
u
rio
sity
o
v
e
r
tec
h
n
o
l
o
g
y
e
a
rn
e
d
h
e
r
a
sc
h
o
lars
h
ip
to
th
e
A
p
p
le
De
v
e
lo
p
e
r
Ac
a
d
e
m
y
in
h
e
r
se
c
o
n
d
y
e
a
r.
W
it
h
a
p
ro
f
o
u
n
d
in
tere
st
in
iOS
d
e
v
e
lo
p
m
e
n
t
a
n
d
m
a
c
h
in
e
lea
rn
in
g
,
s
h
e
wa
s
in
v
o
lv
e
d
i
n
se
v
e
ra
l
p
ro
jec
ts
s
u
c
h
a
s
a
n
a
p
p
li
c
a
ti
o
n
t
h
a
t
lev
e
ra
g
e
s
m
a
c
h
in
e
lea
rn
i
n
g
to
g
e
n
e
ra
te
p
u
b
li
c
sp
e
a
k
in
g
fe
e
d
b
a
c
k
.
He
r
p
a
ss
io
n
li
e
s in
e
x
p
l
o
ri
n
g
re
se
a
rc
h
d
o
m
a
in
s th
a
t
i
n
c
lu
d
e
m
a
c
h
in
e
lea
rn
in
g
,
a
rti
ficia
l
in
tell
ig
e
n
c
e
,
a
n
d
n
a
t
u
ra
l
lan
g
u
a
g
e
p
r
o
c
e
ss
in
g
.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
jev
a
d
d
icte
d
@
g
m
a
il
.
c
o
m
.
Ma
sta
wil
a
Febry
a
n
ti
S
i
m
a
n
ju
n
ta
k
re
c
e
iv
e
d
h
e
r
b
a
c
h
e
lo
r'
s
d
e
g
re
e
in
th
e
In
fo
rm
a
ti
o
n
S
y
ste
m
S
t
u
d
y
P
ro
g
ra
m
,
F
a
c
u
lt
y
o
f
I
n
fo
rm
a
ti
c
s a
n
d
El
e
c
tri
c
a
l
En
g
in
e
e
rin
g
,
I
n
stit
u
t
Tek
n
o
lo
g
i
De
l,
In
d
o
n
e
sia
,
in
2
0
2
4
.
He
r
re
se
a
rc
h
i
n
tere
sts
a
re
n
a
tu
r
a
l
lan
g
u
a
g
e
p
ro
c
e
ss
in
g
,
a
n
d
d
a
ta m
in
in
g
.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
m
a
sta
wilas
ima
n
ju
n
tak
8
2
1
7
1
@
g
m
a
il
.
c
o
m
.
S
a
n
d
r
o
S
in
a
g
a
re
c
e
iv
e
d
h
is
b
a
c
h
e
lo
r'
s
d
e
g
re
e
in
th
e
In
f
o
rm
a
ti
o
n
S
y
ste
m
S
tu
d
y
P
ro
g
ra
m
,
F
a
c
u
lt
y
o
f
In
f
o
rm
a
ti
c
s
a
n
d
El
e
c
tri
c
a
l
En
g
in
e
e
rin
g
,
I
n
stit
u
t
Tek
n
o
l
o
g
i
De
l,
I
n
d
o
n
e
sia
,
in
2
0
2
4
.
His
re
se
a
rc
h
in
tere
sts
a
re
n
a
tu
ra
l
la
n
g
u
a
g
e
p
ro
c
e
ss
in
g
a
n
d
d
a
ta
m
in
i
n
g
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
sa
n
n
d
ro
si
n
a
g
a
8
9
@g
m
a
il
.
c
o
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.