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.
15
,
No
.
5
,
Octo
b
er
2
0
2
6
,
p
p
.
4
3
8
1
~
4
3
9
1
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
15
.i
5
.
p
p
4
3
8
1
-
4
3
9
1
4381
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Stacking
ensembl
e f
o
r mul
tilingua
l
as
pect
-
ba
sed
sent
iment
a
na
ly
sis
of t
o
urism
reviews
B
a
s
wo
ro
Ardi
P
ra
m
o
no
1,
2
,
R
a
hm
a
t
G
er
no
w
o
2
,
Ag
hu
s
So
f
wa
n
3
1
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
mat
i
c
s
E
n
g
i
n
e
e
r
i
n
g
,
F
a
c
u
l
t
y
o
f
I
n
f
o
r
ma
t
i
o
n
a
n
d
C
o
m
m
u
n
i
c
a
t
i
o
n
Te
c
h
n
o
l
o
g
y
,
U
n
i
v
e
r
si
t
a
s Se
maran
g
,
S
e
mara
n
g
,
I
n
d
o
n
e
si
a
2
D
o
c
t
o
r
o
f
I
n
f
o
r
ma
t
i
o
n
S
y
st
e
ms,
G
r
a
d
u
a
t
e
S
c
h
o
o
l
,
U
n
i
v
e
r
si
t
a
s D
i
p
o
n
e
g
o
r
o
,
S
e
mara
n
g
,
I
n
d
o
n
e
si
a
3
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
i
c
a
l
En
g
i
n
e
e
r
i
n
g
,
F
a
c
u
l
t
y
o
f
E
n
g
i
n
e
e
r
i
n
g
,
U
n
i
v
e
r
si
t
a
s
D
i
p
o
n
e
g
o
r
o
,
S
e
mara
n
g
,
I
n
d
o
n
e
s
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
n
7
,
2025
R
ev
is
ed
Au
g
4
,
2
0
2
6
Acc
ep
ted
Au
g
12
,
2
0
2
6
M
u
lt
il
in
g
u
a
l
t
o
u
r
ism
re
v
iew
s
o
ften
e
x
p
re
ss
o
p
in
i
o
n
s
a
b
o
u
t
m
u
lt
ip
le
d
e
stin
a
ti
o
n
a
tt
rib
u
tes
wit
h
i
n
a
s
in
g
le
tex
t
,
m
a
k
i
n
g
a
sp
e
c
t
-
lev
e
l
se
n
ti
m
e
n
t
c
las
sifica
ti
o
n
c
h
a
ll
e
n
g
in
g
.
T
h
is
s
tu
d
y
c
o
m
p
a
re
s
c
las
sic
a
l
m
a
c
h
in
e
lea
rn
in
g
a
n
d
e
n
se
m
b
le
m
e
th
o
d
s
fo
r
m
u
lt
il
a
b
e
l
a
sp
e
c
t
-
b
a
se
d
se
n
t
ime
n
t
a
n
a
ly
sis
o
f
a
m
e
n
it
ies
,
a
tt
ra
c
ti
o
n
s,
a
n
d
a
c
c
e
ss
ib
il
it
y
.
T
h
e
d
a
tas
e
t
c
o
m
p
ris
e
s
1
3
,
0
8
5
In
d
o
n
e
sia
n
-
a
n
d
E
n
g
li
sh
-
lan
g
u
a
g
e
G
o
o
g
le
Re
v
iew
s
lab
e
led
th
r
o
u
g
h
a
ru
le
-
b
a
se
d
p
r
o
c
e
d
u
re
.
Re
v
iew
s
we
re
re
p
re
se
n
ted
u
si
n
g
term
fre
q
u
e
n
c
y
-
in
v
e
rs
e
d
o
c
u
m
e
n
t
f
re
q
u
e
n
c
y
fe
a
tu
re
s
a
n
d
c
las
sified
u
sin
g
lo
g
isti
c
r
e
g
re
ss
io
n
,
m
u
lt
in
o
m
ial
n
a
i
v
e
Ba
y
e
s,
ra
n
d
o
m
fo
re
st
,
so
ft
v
o
ti
n
g
,
a
d
a
p
ti
v
e
b
o
o
stin
g
(
Ad
a
Bo
o
st
)
,
a
n
d
sta
c
k
in
g
u
n
d
e
r
a
one
-
vs
-
re
st
fra
m
e
wo
rk
.
Th
e
sta
c
k
in
g
c
las
sifier
c
o
m
b
in
e
d
lo
g
isti
c
re
g
r
e
ss
io
n
a
n
d
ra
n
d
o
m
fo
re
st
a
s
b
a
s
e
lea
rn
e
rs
with
lo
g
isti
c
re
g
re
ss
io
n
a
s
th
e
m
e
ta
-
lea
rn
e
r.
It
a
c
h
iev
e
d
t
h
e
b
e
st
p
e
rfo
rm
a
n
c
e
a
m
o
n
g
th
e
e
v
a
l
u
a
ted
m
o
d
e
ls,
with
a
m
icro
-
F
1
o
f
0
.
8
6
2
6
,
m
a
c
ro
-
F
1
o
f
0
.
7
2
8
5
,
Ha
m
m
in
g
l
o
ss
o
f
0
.
0
6
3
3
,
a
n
d
su
b
se
t
a
c
c
u
ra
c
y
o
f
0
.
6
8
2
5
.
Th
e
se
fin
d
i
n
g
s
i
n
d
ica
te
t
h
a
t
sta
c
k
i
n
g
e
ff
e
c
ti
v
e
ly
i
n
teg
ra
tes
c
o
m
p
lem
e
n
tary
d
e
c
isio
n
p
a
tt
e
rn
s
a
n
d
p
ro
v
id
e
s
a
re
p
ro
d
u
c
ib
le
fra
m
e
wo
rk
f
o
r
m
u
lt
il
i
n
g
u
a
l
m
u
lt
il
a
b
e
l
a
n
a
ly
sis o
f
to
u
rism
re
v
iew
s.
K
ey
w
o
r
d
s
:
A
m
e
n
i
t
i
e
s
,
a
t
t
r
a
c
t
i
o
n
s
,
a
c
c
e
s
s
i
b
i
l
i
t
y
Asp
ec
t
-
b
ased
s
en
tim
en
t a
n
aly
s
is
Mu
ltil
ab
el
class
if
icatio
n
Mu
ltil
in
g
u
al
to
u
r
is
m
r
e
v
iews
Stack
in
g
en
s
em
b
le
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
:
B
aswo
r
o
Ar
d
i Pr
am
o
n
o
Dep
ar
tm
en
t o
f
I
n
f
o
r
m
atics E
n
g
in
ee
r
in
g
,
Facu
lty
o
f
I
n
f
o
r
m
at
io
n
an
d
C
o
m
m
u
n
icatio
n
T
ec
h
n
o
lo
g
y
Un
iv
er
s
itas
Sem
ar
an
g
St.
So
ek
ar
n
o
-
Hatta,
Sem
ar
a
n
g
5
0
1
9
6
,
C
en
tr
al
J
av
a,
I
n
d
o
n
esia
E
m
ail:
b
aswo
r
o
@
u
s
m
.
ac
.
id
1.
I
NT
RO
D
UCT
I
O
N
Un
d
er
s
tan
d
in
g
tr
a
v
eler
s
'
s
en
ti
m
en
ts
th
r
o
u
g
h
o
n
lin
e
r
e
v
iews
is
in
cr
ea
s
in
g
ly
v
ital
f
o
r
s
u
s
tain
in
g
an
d
en
h
an
cin
g
th
e
co
m
p
etitiv
en
es
s
o
f
to
u
r
is
m
d
esti
n
atio
n
s
g
lo
b
ally
.
User
-
g
en
er
ated
co
n
ten
t
s
ig
n
if
ican
tly
s
h
ap
es
c
o
n
s
u
m
e
r
d
e
c
i
s
i
o
n
s
,
i
m
p
a
ct
i
n
g
t
o
u
r
i
s
m
b
u
s
i
n
es
s
e
s
d
i
r
e
ct
l
y
b
y
i
n
f
l
u
e
n
c
i
n
g
s
e
r
v
i
c
e
r
e
p
u
t
at
i
o
n
a
n
d
q
u
a
l
i
t
y
[
1
]
,
[
2
]
.
T
h
u
s
,
ex
tr
ac
tin
g
p
r
ec
is
e
in
s
ig
h
ts
f
r
o
m
m
u
ltil
in
g
u
al
t
o
u
r
is
m
r
ev
iews b
ec
o
m
es c
r
u
cial
in
th
e
h
ig
h
ly
co
m
p
etitiv
e
in
ter
n
atio
n
al
to
u
r
is
m
m
ar
k
et.
Asp
ec
t
-
b
ased
s
en
tim
en
t
an
a
ly
s
is
(
AB
SA)
,
f
o
cu
s
in
g
ex
p
licitly
o
n
am
en
ities
,
attr
ac
tio
n
s
,
an
d
ac
ce
s
s
ib
ilit
y
,
en
ab
les
a
n
u
a
n
c
ed
u
n
d
er
s
tan
d
in
g
o
f
to
u
r
is
t
e
x
p
er
ien
ce
s
.
Desp
ite
th
ei
r
p
o
ten
tial,
m
u
ltil
in
g
u
al
r
ev
iews
p
o
s
e
d
is
tin
ct
cla
s
s
if
i
ca
tio
n
ch
allen
g
es
d
u
e
to
lan
g
u
ag
e
h
eter
o
g
en
eity
,
g
r
am
m
a
tical
v
ar
iatio
n
,
an
d
cu
ltu
r
al
co
n
tex
t,
n
ec
ess
itatin
g
r
o
b
u
s
t
an
aly
s
is
tech
n
iq
u
es
[
3
]
,
[
4
]
.
Ho
wev
er
,
th
e
s
y
s
tem
atic
ev
alu
atio
n
o
f
class
ical
m
ac
h
in
e
lear
n
in
g
a
n
d
ad
v
an
ce
d
en
s
em
b
le
m
eth
o
d
s
,
esp
ec
ially
s
tack
in
g
,
f
o
r
m
u
ltil
in
g
u
al
AB
SA
r
em
ain
s
lim
ited
,
cr
ea
tin
g
s
ig
n
if
ican
t
k
n
o
wled
g
e
g
ap
s
.
Sp
ec
if
ically
,
co
m
p
r
e
h
en
s
iv
e
co
m
p
a
r
ativ
e
ass
es
s
m
en
ts
ac
r
o
s
s
lan
g
u
ag
es a
n
d
cu
ltu
r
es,
as we
ll a
s
th
e
less
ex
p
lo
r
ed
ac
ce
s
s
ib
ilit
y
asp
ec
t,
ar
e
s
ca
r
ce
[
5
]
–
[
7
]
.
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.
15
,
No
.
5
,
Octo
b
er
2
0
2
6
:
4
3
8
1
-
4
3
9
1
4382
Prio
r
r
esear
ch
p
r
ed
o
m
in
an
tly
u
s
es
class
ica
l
m
ac
h
in
e
lear
n
i
n
g
m
eth
o
d
s
,
s
u
ch
as
lo
g
is
tic
r
eg
r
ess
io
n
,
n
aiv
e
B
ay
es,
an
d
r
an
d
o
m
f
o
r
ests
,
f
o
r
AB
S
A
in
to
u
r
is
m
co
n
tex
ts
,
y
ield
in
g
r
eliab
le
y
e
t
o
f
ten
s
u
b
o
p
tim
al
p
er
f
o
r
m
an
ce
.
T
h
ese
ap
p
r
o
ac
h
es
h
av
e
b
ee
n
r
ep
o
r
ted
f
o
r
h
o
tel
r
ev
iew
AB
SA
an
d
o
th
er
ev
alu
ativ
e
to
u
r
is
m
co
n
ten
t
[
8
]
.
Ho
wev
er
,
m
u
ltil
in
g
u
al
s
ettin
g
s
in
tr
o
d
u
ce
ad
d
itio
n
al
lin
g
u
is
tic
v
ar
iab
ilit
y
an
d
la
b
el
am
b
ig
u
ity
th
at
r
em
ain
in
s
u
f
f
icien
tly
ad
d
r
ess
ed
in
m
an
y
ex
is
tin
g
s
tu
d
ies
[
9
]
,
[
1
0
]
.
T
h
e
liter
atu
r
e
f
u
r
th
e
r
in
d
ica
tes
an
u
n
d
e
r
r
ep
r
esen
tatio
n
o
f
en
s
em
b
le
m
eth
o
d
s
,
s
u
ch
as
v
o
tin
g
,
ad
ap
tiv
e
b
o
o
s
tin
g
(
Ad
aBo
o
s
t
)
,
an
d
s
tack
in
g
,
in
AB
SA
s
t
u
d
ies,
d
esp
ite
th
eir
r
ec
o
g
n
iz
ed
ad
v
an
ta
g
es
f
o
r
im
p
r
o
v
in
g
p
r
ed
ictiv
e
ac
cu
r
ac
y
b
y
co
m
b
in
in
g
th
e
s
tr
en
g
th
s
o
f
m
u
ltip
le
class
if
ier
s
[
1
1
]
.
R
ec
en
t stu
d
ies co
n
f
ir
m
th
at
en
s
em
b
le
ap
p
r
o
ac
h
es
ca
n
s
ig
n
if
ican
tly
im
p
r
o
v
e
class
if
icatio
n
r
o
b
u
s
tn
ess
,
es
p
ec
ially
cr
u
cial
in
m
u
ltil
in
g
u
al
an
d
cu
ltu
r
ally
d
iv
er
s
e
d
ata
[
1
2
]
,
[
1
3
]
.
M
o
r
e
o
v
er
,
e
x
is
tin
g
r
esear
ch
in
ad
e
q
u
ately
ad
d
r
ess
es
ac
ce
s
s
ib
ilit
y
with
in
th
e
am
en
ities
,
attr
ac
tio
n
s
,
an
d
ac
ce
s
s
ib
i
lity
f
r
am
ewo
r
k
,
t
h
er
eb
y
lim
itin
g
co
m
p
r
eh
en
s
iv
e
s
en
tim
en
t
an
aly
s
is
in
to
u
r
is
m
r
ev
iews.
Acc
ess
ib
ilit
y
s
ig
n
if
ican
tly
im
p
ac
ts
tr
av
eler
s
atis
f
ac
tio
n
,
p
ar
ticu
lar
ly
f
o
r
t
h
o
s
e
w
i
t
h
d
is
a
b
il
i
ti
e
s
,
y
et
th
i
s
c
r
u
c
i
al
c
o
m
p
o
n
e
n
t
i
s
f
r
e
q
u
en
t
l
y
n
e
g
l
e
c
t
e
d
o
r
s
u
p
e
r
f
ic
i
a
ll
y
e
x
a
m
i
n
e
d
[
6
]
,
[
1
4
]
.
T
h
is
s
tu
d
y
h
y
p
o
th
esizes
th
at
a
s
tack
in
g
en
s
em
b
le
ap
p
r
o
ac
h
ca
n
s
ig
n
if
ican
tly
en
h
an
ce
m
u
ltil
in
g
u
al
AB
SA
ac
cu
r
ac
y
co
m
p
ar
ed
to
class
ical
m
o
d
els
an
d
s
im
p
ler
en
s
em
b
le
m
eth
o
d
s
.
T
h
u
s
,
th
i
s
r
esear
ch
aim
s
to
co
m
p
ar
ativ
ely
ev
alu
ate
t
h
ese
m
eth
o
d
s
u
s
in
g
a
d
ataset
o
f
I
n
d
o
n
esian
an
d
E
n
g
lis
h
r
ev
iews
f
r
o
m
C
en
tr
al
J
av
a,
f
o
cu
s
in
g
s
p
ec
if
ically
o
n
t
h
e
t
h
r
ee
asp
ec
ts
(
am
en
ities
,
attr
ac
tio
n
s
,
an
d
ac
ce
s
s
ib
ilit
y
)
.
B
y
a
d
d
r
ess
in
g
id
en
tifie
d
k
n
o
wled
g
e
g
ap
s
,
th
e
r
esear
c
h
co
n
tr
ib
u
tes
m
eth
o
d
o
l
o
g
ical
clar
ity
an
d
im
p
r
o
v
ed
p
r
ed
i
ctiv
e
f
r
am
ewo
r
k
s
,
u
ltima
tely
aid
in
g
d
esti
n
atio
n
m
an
ag
er
s
an
d
d
ev
el
o
p
er
s
in
r
e
f
in
in
g
to
u
r
is
m
r
ec
o
m
m
en
d
atio
n
s
y
s
tem
s
.
2.
M
E
T
H
O
D
T
h
is
s
tu
d
y
a
d
o
p
ts
a
s
tr
u
ctu
r
e
d
an
d
th
o
r
o
u
g
h
co
m
p
u
tatio
n
a
l
an
d
ex
p
e
r
im
en
tal
d
esig
n
to
co
n
d
u
ct
a
co
m
p
ar
ativ
e
e
v
alu
atio
n
an
d
a
n
aly
ze
th
e
p
er
f
o
r
m
an
ce
o
f
a
s
et
o
f
m
ac
h
in
e
lear
n
in
g
m
o
d
els
.
T
h
e
m
o
d
els
test
ed
in
clu
d
e
class
ical
ca
teg
o
r
y
an
d
en
s
em
b
le
ap
p
r
o
ac
h
es
ap
p
lied
to
th
e
AB
SA
ta
s
k
in
a
m
u
ltil
i
n
g
u
al
co
n
te
x
t.
T
h
is
in
v
esti
g
atio
n
f
o
cu
s
es
o
n
an
al
y
zin
g
d
ig
ital
to
u
r
is
m
r
ev
iews,
with
an
em
p
h
asis
o
n
th
r
ee
cr
u
cial
asp
ec
ts
in
th
e
to
u
r
is
m
d
o
m
ain
:
am
en
ity
,
attr
ac
tio
n
,
an
d
ac
ce
s
s
ib
ilit
y
.
T
h
e
s
et
o
f
m
eth
o
d
o
lo
g
ies im
p
lem
e
n
ted
in
th
is
r
esear
ch
in
clu
d
es
s
ev
er
al
f
u
n
d
am
en
tal
s
tag
es
th
at
ar
e
in
ter
co
n
n
ec
te
d
,
s
tar
tin
g
f
r
o
m
th
e
ac
q
u
is
itio
n
an
d
lab
elin
g
o
f
r
ev
iew
d
ata,
f
o
llo
wed
b
y
te
x
t
p
r
e
p
r
o
ce
s
s
in
g
f
o
r
n
o
r
m
ali
za
tio
n
an
d
clea
n
i
n
g
,
ex
tr
ac
ti
o
n
o
f
i
n
f
o
r
m
ativ
e
f
ea
tu
r
es
f
r
o
m
tex
t,
tr
an
s
f
o
r
m
a
tio
n
o
f
lab
el
f
o
r
m
at
f
o
r
co
m
p
atib
ilit
y
with
m
u
ltil
ab
el
clas
s
i
f
icatio
n
alg
o
r
ith
m
s
,
s
e
g
m
e
n
t
a
ti
o
n
o
f
d
a
t
a
s
et
s
i
n
t
o
tr
a
i
n
i
n
g
a
n
d
t
e
s
t
d
a
t
as
e
ts
,
i
m
p
l
em
e
n
t
a
t
i
o
n
o
f
m
u
l
t
il
a
b
e
l
cl
a
s
s
i
f
ic
a
t
i
o
n
a
r
c
h
i
te
c
t
u
r
e
,
m
o
d
e
l
t
r
a
i
n
i
n
g
p
r
o
c
e
s
s
,
e
v
a
l
u
at
i
o
n
o
f
m
o
d
e
l
p
e
r
f
o
r
m
a
n
c
e
u
s
i
n
g
s
t
a
n
d
a
r
d
m
e
t
r
i
c
s
,
t
o
t
h
e
f
i
n
a
l
s
tag
e
o
f
m
o
d
el
s
elec
tio
n
with
o
p
tim
al
p
er
f
o
r
m
an
ce
.
T
h
e
o
v
er
all
m
et
h
o
d
o
lo
g
ical
wo
r
k
f
lo
w
is
p
r
esen
te
d
in
Fig
u
r
e
1
.
Fig
u
r
e
1
.
Ov
e
r
all
wo
r
k
f
l
o
w
o
f
th
e
p
r
o
p
o
s
ed
r
esear
c
h
m
eth
o
d
2
.
1
.
Da
t
a
c
o
llect
io
n a
nd
f
ilte
ring
T
h
e
p
r
im
a
r
y
d
ata
s
o
u
r
ce
f
o
r
th
is
s
tu
d
y
is
u
s
er
r
ev
iews
co
lle
cted
f
r
o
m
th
e
Go
o
g
le
R
ev
iew
s
p
latf
o
r
m
.
T
h
is
p
latf
o
r
m
was
ch
o
s
en
b
ec
au
s
e
it
p
r
o
v
id
es
a
lar
g
e,
d
iv
e
r
s
e
v
o
lu
m
e
o
f
r
e
v
iew
d
ata
co
v
er
in
g
a
wid
e
r
an
g
e
o
f
to
u
r
is
m
d
esti
n
atio
n
s
an
d
is
wr
itten
in
m
u
ltip
le
lan
g
u
ag
es
b
y
tr
av
eler
s
with
d
if
f
er
en
t
b
ac
k
g
r
o
u
n
d
s
.
Sp
ec
if
ically
,
th
e
d
ataset
co
m
p
r
is
es
1
3
,
0
8
5
r
ev
iews
f
r
o
m
t
en
p
o
p
u
lar
to
u
r
is
m
d
esti
n
atio
n
s
in
C
en
tr
al
J
av
a,
I
n
d
o
n
esia.
T
h
e
s
elec
tio
n
o
f
th
ese
d
esti
n
atio
n
s
aim
s
to
o
b
tain
a
r
ep
r
esen
tativ
e
v
ar
iet
y
o
f
r
ev
iews
f
r
o
m
d
if
f
er
en
t ty
p
es o
f
to
u
r
is
t a
ttra
ctio
n
s
.
I
n
clu
s
io
n
cr
iter
ia
wer
e
s
tr
ictly
ap
p
lied
d
u
r
i
n
g
th
e
d
ata
c
o
llectio
n
p
r
o
ce
s
s
to
en
s
u
r
e
th
e
r
ele
v
an
ce
an
d
q
u
ality
o
f
t
h
e
d
ata
to
b
e
an
al
y
ze
d
:
i)
R
ev
iew
lan
g
u
ag
e:
o
n
l
y
r
ev
i
ews
wr
itten
in
I
n
d
o
n
esian
an
d
E
n
g
lis
h
lan
g
u
ag
e
wer
e
i
n
clu
d
ed
.
T
h
is
d
ec
is
io
n
was b
ased
o
n
th
e
r
ese
ar
ch
'
s
m
u
ltil
in
g
u
al
f
o
cu
s
an
d
t
h
e
d
o
m
in
a
n
ce
o
f
th
ese
two
lan
g
u
ag
es in
th
e
co
n
tex
t o
f
d
ig
ital to
u
r
is
m
in
I
n
d
o
n
esia.
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
S
ta
ck
in
g
e
n
s
emb
le
fo
r
mu
ltil
in
g
u
a
l a
s
p
ec
t
-
b
a
s
ed
s
en
timen
t a
n
a
lysi
s
o
f to
u
r
is
m
…
(
B
a
s
w
o
r
o
A
r
d
i P
r
a
mo
n
o
)
4383
ii)
R
ev
iew
len
g
th
:
ea
ch
r
ev
iew
m
u
s
t
b
e
at
least
5
0
w
o
r
d
s
lo
n
g
.
T
h
ese
r
estrictio
n
s
ar
e
ap
p
lied
to
en
s
u
r
e
th
a
t
th
e
r
ev
iew
co
n
tain
s
s
u
f
f
icien
t
co
n
tex
tu
al
i
n
f
o
r
m
atio
n
f
o
r
in
-
d
ep
th
a
n
aly
s
is
o
f
asp
ec
ts
an
d
s
en
tim
en
t
an
d
to
av
o
id
r
ev
iews th
at
ar
e
to
o
s
h
o
r
t a
n
d
u
n
i
n
f
o
r
m
ativ
e.
iii)
R
elev
an
ce
o
f
asp
ec
ts
:
th
e
r
e
v
iew
m
u
s
t
ex
p
licitly
m
en
tio
n
at
least
o
n
e
o
f
th
e
th
r
ee
m
ain
as
p
ec
ts
th
at
th
e
r
esear
ch
f
o
cu
s
es
o
n
:
am
en
it
ies
(
f
ac
ilit
ies
an
d
s
er
v
ices)
,
attr
ac
tio
n
s
(
th
e
m
ain
attr
ac
tio
n
o
f
th
e
d
esti
n
atio
n
)
,
o
r
ac
ce
s
s
ib
ilit
y
(
ea
s
e
o
f
r
ea
ch
in
g
an
d
m
o
v
i
n
g
ar
o
u
n
d
t
h
e
d
esti
n
atio
n
)
.
Af
ter
an
in
itial
co
llectio
n
a
n
d
s
cr
ee
n
in
g
p
r
o
ce
s
s
b
ased
o
n
t
h
e
ab
o
v
e
cr
iter
ia,
r
ev
iews
th
a
t
m
ee
t
th
e
cr
iter
ia
ar
e
th
en
an
aly
ze
d
at
th
e
clau
s
e
lev
el
u
s
in
g
a
r
u
le
-
b
ased
lab
elin
g
a
p
p
r
o
ac
h
,
with
th
e
h
elp
o
f
a
s
tr
u
ctu
r
ed
k
ey
w
o
r
d
d
ictio
n
ar
y
to
id
en
tify
a
p
p
r
o
p
r
iate
asp
ec
ts
an
d
s
en
tim
en
t
p
o
lar
ities
(
p
o
s
itiv
e,
n
eg
ativ
e,
n
eu
tr
al)
[
3
]
.
T
h
e
d
ataset
u
s
ed
in
th
is
s
tu
d
y
is
p
u
b
licly
av
ailab
le
o
n
Z
en
o
d
o
an
d
co
n
tain
s
lab
eled
to
u
r
is
m
r
e
v
iews
f
r
o
m
C
en
tr
a
l
J
av
a,
I
n
d
o
n
esia,
an
n
o
tated
u
s
in
g
a
r
u
le
-
b
ased
ap
p
r
o
ac
h
[
1
5
]
.
T
h
e
d
ataset
in
clu
d
es
m
u
ltil
in
g
u
al
r
ev
iew
tex
ts
an
d
th
e
co
r
r
esp
o
n
d
i
n
g
asp
ec
t
s
en
tim
en
t
lab
els
u
s
e
d
in
th
e
m
u
ltil
ab
el
class
if
icatio
n
ex
p
er
im
en
ts
.
2
.
2
.
T
ex
t
prepro
ce
s
s
ing
B
ef
o
r
e
f
ea
tu
r
e
ex
tr
ac
tio
n
,
th
e
m
u
ltil
in
g
u
al
r
ev
iew
tex
t
(
s
to
r
ed
in
th
e
r
ev
iew_
clea
n
co
lu
m
n
af
ter
th
e
in
itial
clea
n
u
p
d
u
r
in
g
d
ata
c
o
llectio
n
)
u
n
d
er
g
o
es
s
ev
er
al
s
tan
d
ar
d
p
r
ep
r
o
ce
s
s
in
g
s
tep
s
to
n
o
r
m
alize
f
u
r
th
e
r
an
d
clea
n
it.
T
h
e
m
u
ltil
in
g
u
al
r
ev
iew
tex
t
p
r
ep
r
o
ce
s
s
in
g
p
r
o
ce
s
s
in
clu
d
es
s
ev
er
al
im
p
o
r
tan
t
s
tag
es:
n
o
r
m
aliza
tio
n
,
t
o
k
en
izatio
n
,
an
d
n
o
is
e
r
em
o
v
al.
No
r
m
aliz
atio
n
en
s
u
r
es
co
n
s
is
ten
cy
in
tex
t
p
r
o
ce
s
s
in
g
b
y
co
n
v
er
tin
g
en
tire
ch
ar
ac
ter
s
to
lo
wer
ca
s
e
an
d
h
an
d
lin
g
s
p
ellin
g
v
ar
iatio
n
s
.
T
o
k
en
izatio
n
b
r
ea
k
s
d
o
wn
tex
t
in
to
s
m
aller
u
n
its
(
to
k
en
s
)
,
w
h
ich
is
esp
ec
ially
im
p
o
r
tan
t
f
o
r
lan
g
u
ag
es
lik
e
I
n
d
o
n
esian
an
d
E
n
g
lis
h
b
ec
au
s
e
o
f
th
ei
r
d
i
f
f
er
en
t
s
y
n
tactic
s
tr
u
ctu
r
es.
No
is
e
r
em
o
v
al
elim
in
ates
ir
r
elev
an
t
elem
e
n
ts
,
s
u
ch
as
p
u
n
ctu
atio
n
a
n
d
s
to
p
wo
r
d
s
,
th
at
ca
n
o
b
s
cu
r
e
t
h
e
s
en
tim
en
t
in
u
s
er
-
g
en
e
r
ate
d
co
n
ten
t
[
1
6
]
.
T
h
is
s
tag
e
is
c
r
u
cial
f
o
r
r
ed
u
cin
g
n
o
is
e
an
d
im
p
r
o
v
i
n
g
th
e
q
u
alit
y
o
f
th
e
f
ea
tu
r
es to
b
e
ex
tr
ac
te
d
f
o
r
m
o
d
elin
g
.
2
.
3
.
L
a
belin
g
s
chem
e
a
nd
la
bel t
ra
ns
f
o
rm
a
t
io
n
L
ab
elin
g
s
ch
em
e
a
n
d
la
b
el
tr
an
s
f
o
r
m
atio
n
.
T
h
e
d
ataset
was
an
n
o
tated
u
s
in
g
a
r
u
le
-
b
ase
d
s
ch
em
e
s
u
p
p
o
r
ted
b
y
a
s
tr
u
ctu
r
ed
b
ili
n
g
u
al
k
ey
wo
r
d
d
ictio
n
ar
y
f
o
r
th
e
th
r
ee
to
u
r
is
m
asp
ec
ts
(
am
en
ities
,
attr
ac
tio
n
s
,
an
d
ac
ce
s
s
ib
ilit
y
)
an
d
th
r
ee
s
e
n
tim
en
t
p
o
lar
ities
(
p
o
s
itiv
e,
n
e
g
ativ
e,
an
d
n
eu
tr
al)
.
E
ac
h
r
ev
i
ew
was
s
eg
m
en
ted
in
to
clau
s
es
to
r
ed
u
ce
s
en
tim
en
t
m
ix
in
g
ac
r
o
s
s
asp
ec
ts
,
an
d
ea
ch
clau
s
e
was
m
atch
ed
ag
ain
s
t
th
e
asp
ec
t
d
ictio
n
ar
y
t
o
id
en
tif
y
th
e
r
elev
an
t
asp
ec
t
ca
teg
o
r
y
.
Sen
ti
m
en
t
p
o
lar
ity
was
th
en
ass
ig
n
ed
u
s
in
g
p
o
lar
ity
k
ey
wo
r
d
s
an
d
co
n
tex
tu
al
r
u
les (
e.
g
.
,
n
eg
atio
n
h
an
d
lin
g
)
,
p
r
o
d
u
cin
g
cla
u
s
e
-
lev
el
asp
ec
t sen
tim
en
t p
air
s
.
T
o
s
u
p
p
o
r
t m
u
ltil
ab
el
lear
n
in
g
at
th
e
r
ev
iew
lev
el,
all
c
lau
s
e
-
lev
el
p
air
s
with
in
th
e
s
am
e
r
ev
iew
wer
e
ag
g
r
eg
ated
in
to
a
s
et
o
f
co
m
p
o
s
ite
lab
els
in
th
e
f
o
r
m
at
"
a
s
p
ec
t_
p
o
lar
ity
".
T
h
er
ef
o
r
e,
o
n
e
r
ev
iew
ca
n
h
av
e
m
u
ltip
le
lab
els
s
im
u
ltan
eo
u
s
ly
(
e.
g
.
,
am
e
n
ity
_
p
o
s
itiv
e
an
d
ac
ce
s
s
ib
ilit
y
_
n
eg
ativ
e
)
.
T
h
e
f
in
a
l
lab
el
s
et
co
n
s
is
ts
o
f
3
asp
ec
ts
×
3
p
o
lar
ities
,
f
o
r
a
to
tal
o
f
9
p
o
s
s
ib
le
lab
el
s
.
T
h
e
ag
g
r
eg
ated
la
b
el
s
ets
wer
e
co
n
v
e
r
ted
to
a
m
u
lti
-
h
o
t
r
e
p
r
esen
tatio
n
u
s
in
g
a
Mu
ltiLab
elB
in
ar
izer
to
c
o
n
s
tr
u
ct
th
e
b
i
n
ar
y
ta
r
g
et
m
a
tr
ix
f
o
r
one
-
vs
-
r
est
(
Ov
R
)
tr
ain
in
g
a
n
d
ev
al
u
atio
n
.
2
.
4
.
F
e
a
t
ures e
x
t
ra
c
t
io
n
Fo
r
f
ea
tu
r
e
ex
tr
ac
tio
n
,
a
ter
m
f
r
eq
u
e
n
cy
-
in
v
er
s
e
d
o
cu
m
en
t
f
r
eq
u
en
c
y
(
TF
-
I
DF
)
v
ec
to
r
izer
is
u
s
ed
,
a
co
m
m
o
n
tec
h
n
iq
u
e
i
n
m
u
ltil
in
g
u
al
s
en
tim
en
t
an
aly
s
is
[
3
]
.
TF
-
I
DF
h
ig
h
lig
h
ts
th
e
im
p
o
r
ta
n
ce
o
f
ea
ch
wo
r
d
in
th
e
co
n
tex
t
o
f
th
e
e
n
tire
co
r
p
u
s
an
d
ad
ap
ts
it
to
th
e
f
r
eq
u
en
c
y
o
f
co
m
m
o
n
wo
r
d
s
,
en
s
u
r
in
g
th
at
m
o
r
e
r
elev
an
t
f
ea
tu
r
es
ar
e
g
iv
en
a
p
p
r
o
p
r
iate
weig
h
t
in
th
e
class
if
icatio
n
.
T
f
id
f
Vec
to
r
izer
f
r
o
m
th
e
s
cik
it
-
lear
n
lib
r
ar
y
was
u
s
ed
to
im
p
lem
en
t
th
is
r
esear
ch
co
d
e.
T
h
e
m
ax
_
f
ea
tu
r
e
s
=
5
0
0
0
p
a
r
am
eter
is
ap
p
l
ied
,
m
ea
n
in
g
o
n
l
y
5
0
0
0
w
o
r
d
s
with
th
e
h
i
g
h
es
t
T
F
-
I
DF
s
co
r
e
f
r
o
m
th
e
e
n
tire
co
r
p
u
s
will
b
e
co
n
s
id
e
r
ed
f
ea
tu
r
es.
T
h
is
m
ax
_
f
ea
tu
r
es
also
s
er
v
es
as
a
d
im
en
s
io
n
ality
r
ed
u
ctio
n
m
e
th
o
d
,
h
elp
in
g
r
ed
u
ce
m
o
d
el
c
o
m
p
lex
ity
an
d
th
e
r
is
k
o
f
o
v
er
f
itti
n
g
.
T
h
e
r
esu
lt
o
f
th
is
f
ea
tu
r
e
ex
tr
ac
tio
n
s
tag
e
is
a
n
u
m
e
r
ical
f
ea
tu
r
e
m
at
r
ix
,
d
e
n
o
ted
X.
E
ac
h
r
o
w
o
f
X
c
o
r
r
esp
o
n
d
s
to
a
r
ev
i
ew,
an
d
ea
ch
co
lu
m
n
co
r
r
esp
o
n
d
s
to
a
wo
r
d
f
ea
tu
r
e
with
its
T
F
-
I
DF w
eig
h
t.
2
.
5
.
L
a
bel t
ra
ns
f
o
r
m
a
t
io
n
a
nd
da
t
a
s
pli
t
t
ing
T
h
e
s
en
tim
en
t
-
asp
ec
t
lab
els
r
esu
ltin
g
f
r
o
m
th
e
lab
elin
g
p
r
o
ce
s
s
(
as
a
lis
t
o
f
tu
p
les)
ar
e
tr
an
s
f
o
r
m
ed
in
to
a
m
u
ltil
ab
el
b
i
n
ar
y
f
o
r
m
a
t
u
s
in
g
t
h
e
Mu
ltiLab
elB
in
ar
iz
er
f
r
o
m
s
cik
it
-
lear
n
.
E
ac
h
u
n
i
q
u
e
c
o
m
b
in
atio
n
o
f
s
en
tim
en
t
asp
ec
ts
(
e.
g
.
,
'
am
en
itas
_
p
o
s
itiv
e'
)
b
ec
o
m
es
a
s
i
n
g
le
b
i
n
ar
y
tar
g
et
c
o
lu
m
n
.
T
h
e
d
atasets
,
wh
ich
a
l
r
e
a
d
y
i
n
c
l
u
d
e
d
n
u
m
e
r
i
c
a
l
f
e
at
u
r
e
s
(
X
)
a
n
d
b
i
n
a
r
y
l
a
b
e
l
s
(
y
)
,
w
e
r
e
t
h
e
n
s
p
l
i
t
a
t
a
n
8
0
:
2
0
r
a
t
i
o
f
o
r
t
r
a
i
n
i
n
g
a
n
d
t
e
s
t
d
at
a
.
T
h
is
r
a
t
i
o
w
a
s
c
h
o
s
e
n
t
o
b
a
l
a
n
c
e
s
u
f
f
ic
i
e
n
t
le
a
r
n
i
n
g
an
d
r
e
p
r
e
s
e
n
t
a
t
i
v
e
e
v
a
l
u
a
t
i
o
n
o
f
u
n
s
e
e
n
d
a
t
a
[
9
]
.
2
.
6
.
M
ultila
bel c
la
s
s
if
ica
t
io
n f
ra
m
ewo
r
k
a
nd
t
este
d m
o
de
ls
T
h
is
s
tu
d
y
f
o
r
m
u
lates
m
u
ltil
in
g
u
al
AB
SA
as
a
m
u
lti
lab
el
te
x
t
class
if
icatio
n
p
r
o
b
lem
,
wh
e
r
e
a
s
in
g
le
r
ev
iew
m
ay
co
n
tain
m
u
ltip
le
asp
ec
t
s
en
tim
en
t
p
air
s
.
E
ac
h
r
ev
iew
is
r
ep
r
esen
ted
b
y
a
s
et
o
f
co
m
p
o
s
ite
lab
els
f
o
r
m
ed
b
y
co
m
b
in
in
g
th
e
a
s
p
ec
t
ca
teg
o
r
y
an
d
s
en
tim
en
t
p
o
lar
ity
(
e.
g
.
,
"
asp
ec
t_
p
o
la
r
ity
")
,
r
esu
ltin
g
in
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.
15
,
No
.
5
,
Octo
b
er
2
0
2
6
:
4
3
8
1
-
4
3
9
1
4384
3
asp
ec
ts
×
3
p
o
lar
ities
=
9
p
o
s
s
ib
le
lab
els.
T
h
e
lab
el
s
ets
ar
e
b
in
ar
ized
u
s
in
g
Mu
ltiLab
el
B
in
ar
izer
,
p
r
o
d
u
cin
g
a
m
u
lti
-
h
o
t ta
r
g
et
v
ec
to
r
f
o
r
ea
ch
r
ev
iew.
Fo
r
tex
tu
al
r
e
p
r
esen
tatio
n
,
T
F
-
I
DF
v
ec
to
r
izatio
n
is
u
s
ed
o
n
th
e
clea
n
ed
r
ev
iew
tex
t
(
r
ev
i
ew_
clea
n
)
with
m
ax
_
f
ea
tu
r
es
=
5
0
0
0
(
d
ef
au
lt
u
n
ig
r
a
m
s
ettin
g
)
.
T
h
e
d
ataset
is
s
p
lit
in
to
tr
ain
in
g
an
d
test
s
ets
u
s
in
g
an
8
0
/2
0
r
atio
with
a
f
ix
ed
r
an
d
o
m
s
ee
d
(
r
an
d
o
m
_
s
tate
=
4
2
)
.
B
ec
au
s
e
m
u
ltil
ab
el
lear
n
in
g
r
eq
u
ir
es
m
o
d
elin
g
ea
ch
lab
el
as
a
b
in
ar
y
d
ec
is
io
n
,
we
em
p
lo
y
a
O
v
R
s
tr
ateg
y
a
s
a
wr
ap
p
er
f
o
r
all
class
if
ier
s
,
en
ab
lin
g
c
o
n
s
is
ten
t
co
m
p
ar
is
o
n
ac
r
o
s
s
b
aselin
e
an
d
en
s
em
b
le
ap
p
r
o
ac
h
es.
i)
B
aselin
e
m
o
d
els
:
‒
L
o
g
is
tic
r
eg
r
ess
io
n
(
Ov
R
)
: L
o
g
is
ticR
eg
r
ess
io
n
(
m
ax
_
iter
=
1
0
0
0
,
r
an
d
o
m
_
s
tate
=
4
2
)
.
‒
Mu
ltin
o
m
ial
n
aiv
e
B
ay
es
(
Ov
R
)
:
Mu
ltin
o
m
ialNB
(
)
as
a
p
r
o
b
ab
ilis
tic
b
aselin
e
s
u
itab
le
f
o
r
T
F
-
I
DF
f
ea
tu
r
es.
‒
R
an
d
o
m
f
o
r
est
(
Ov
R
)
: Ran
d
o
m
Fo
r
estC
las
s
if
ier
(
n
_
esti
m
ato
r
s
=
1
0
0
,
r
a
n
d
o
m
_
s
tate
=
4
2
)
.
ii)
E
n
s
em
b
le
m
o
d
els
:
‒
Vo
tin
g
(
O
v
R
,
s
o
f
t
V
o
tin
g
)
:
Vo
tin
g
C
lass
if
ier
(
v
o
tin
g
=
'
s
o
f
t'
)
with
co
m
p
o
n
en
t
lear
n
e
r
s
lo
g
is
tic
r
eg
r
ess
io
n
,
m
u
ltin
o
m
ial
n
aiv
e
B
ay
es,
an
d
r
an
d
o
m
f
o
r
es
t
.
So
f
t
v
o
tin
g
ag
g
r
eg
ates
p
r
ed
icted
p
r
o
b
a
b
ilit
ies to
im
p
r
o
v
e
r
o
b
u
s
tn
ess
.
‒
Ad
aBo
o
s
t
(
Ov
R
)
:
Ad
aBo
o
s
tC
las
s
if
ier
(
esti
m
ato
r
=
L
o
g
is
ticR
eg
r
ess
io
n
(
m
ax
_
iter
=
1
0
0
0
,
r
an
d
o
m
_
s
tate
=
4
2
)
,
n
_
esti
m
a
to
r
s
=
5
0
,
r
an
d
o
m
_
s
tate
=
4
2
)
to
iter
ativ
ely
em
p
h
asize
m
is
class
if
ied
s
am
p
les.
‒
Stack
in
g
(
Ov
R
)
:
Stack
in
g
C
la
s
s
if
ier
with
b
ase
lear
n
er
s
lo
g
i
s
tic
r
eg
r
ess
io
n
an
d
r
an
d
o
m
f
o
r
est
,
an
d
lo
g
is
tic
r
eg
r
ess
io
n
as
th
e
m
eta
-
lear
n
er
(
f
in
al_
esti
m
ato
r
=
L
o
g
is
ticR
eg
r
ess
io
n
(
)
)
.
W
e
s
et
p
ass
th
r
o
u
g
h
=
t
r
u
e
s
o
m
eta
-
le
ar
n
er
ca
n
ac
ce
s
s
b
o
th
th
e
o
r
ig
in
al
T
F
-
I
DF
f
ea
tu
r
es
an
d
b
ase
-
m
o
d
el
p
r
ed
ictio
n
s
.
Fig
u
r
e
2
illu
s
tr
a
tes
en
d
-
to
-
en
d
e
x
p
er
im
e
n
tal
s
etu
p
u
s
ed
in
th
is
s
tu
d
y
,
in
clu
d
in
g
p
r
ep
r
o
ce
s
s
in
g
,
lab
el
co
n
s
tr
u
cti
o
n
,
f
ea
tu
r
e
ex
tr
ac
tio
n
,
m
o
d
el
t
r
ain
in
g
,
a
n
d
m
u
ltil
ab
el
ev
alu
at
io
n
.
T
h
e
wo
r
k
f
lo
w
s
tar
ts
f
r
o
m
th
e
Z
en
o
d
o
d
ataset
an
d
p
r
o
ce
e
d
s
with
tex
t
p
r
ep
r
o
ce
s
s
in
g
to
p
r
o
d
u
ce
clea
n
ed
r
e
v
iew
tex
ts
.
Fro
m
t
h
e
p
r
e
p
r
o
ce
s
s
ed
d
ata,
two
p
ar
allel
p
r
o
ce
s
s
es
ar
e
p
er
f
o
r
m
ed
:
i
)
a
r
u
le
-
b
ased
lab
elin
g
p
r
o
ce
d
u
r
e
g
e
n
er
ates
r
ev
iew
-
lev
el
m
u
ltil
ab
el
tar
g
ets,
wh
ich
ar
e
tr
an
s
f
o
r
m
ed
in
to
a
m
u
lti
-
h
o
t
r
ep
r
esen
tatio
n
u
s
in
g
Mu
ltiLab
elB
in
ar
izer
,
an
d
ii)
T
F
–
I
DF
v
ec
to
r
izatio
n
(
m
ax
_
f
ea
tu
r
es
=
5
0
0
0
)
co
n
s
tr
u
cts
th
e
f
ea
tu
r
e
m
atr
ix
.
T
h
e
d
ata
ar
e
th
en
s
p
lit
in
to
tr
ain
in
g
an
d
test
in
g
s
ets
(
8
0
/2
0
,
r
an
d
o
m
_
s
tate
=
4
2
)
an
d
tr
ain
ed
u
n
d
er
a
Ov
R
m
u
ltil
ab
el
wr
a
p
p
er
to
e
n
ab
le
a
f
air
c
o
m
p
ar
is
o
n
b
etwe
en
b
aselin
e
m
o
d
els
(
lo
g
is
tic
r
eg
r
ess
io
n
,
m
u
ltin
o
m
ial
n
aiv
e
B
ay
es,
r
a
n
d
o
m
f
o
r
est)
an
d
en
s
em
b
le
ap
p
r
o
ac
h
es
(
v
o
tin
g
,
Ad
aBo
o
s
t,
an
d
s
tack
in
g
)
.
Fin
ally
,
m
o
d
el
p
er
f
o
r
m
a
n
ce
i
s
ev
alu
ated
u
s
in
g
m
icr
o
-
an
d
m
ac
r
o
-
F1
,
p
r
ec
is
io
n
,
r
ec
all,
Ham
m
in
g
lo
s
s
,
a
n
d
s
u
b
s
et
ac
cu
r
ac
y
,
an
d
th
e
r
esu
ltin
g
m
etr
ics ar
e
co
m
p
iled
in
t
h
e
r
ep
o
r
te
d
r
esu
lts
tab
les.
Fig
u
r
e
2
.
E
x
p
er
im
e
n
tal
s
etu
p
o
f
th
e
m
u
ltil
in
g
u
al
m
u
ltil
ab
el
AB
SA p
ip
elin
e
2
.
7
.
E
v
a
lua
t
i
o
n
m
et
rics
T
h
e
ev
alu
atio
n
u
s
ed
f
iv
e
m
etr
ics:
F1
-
s
co
r
e
(
m
icr
o
an
d
m
ac
r
o
)
,
p
r
ec
is
io
n
,
r
ec
all,
H
am
m
in
g
lo
s
s
,
an
d
s
u
b
s
et
ac
cu
r
ac
y
.
T
h
is
m
etr
ic
p
r
o
v
id
es
a
co
m
p
r
eh
e
n
s
iv
e
o
v
er
v
iew
o
f
t
h
e
class
if
icatio
n
p
er
f
o
r
m
an
ce
in
m
u
ltil
ab
el
task
s
.
T
h
e
F1
-
s
co
r
e
co
m
b
in
es p
r
ec
is
io
n
an
d
r
ec
all
,
wh
ile
Ham
m
in
g
lo
s
s
m
ea
s
u
r
es p
r
ed
ictio
n
er
r
o
r
s
p
er
lab
el,
a
n
d
s
u
b
s
et
ac
cu
r
a
cy
ass
ess
e
s
o
v
er
all
lab
el
s
u
i
tab
ilit
y
[
1
7
]
.
Mic
r
o
a
n
d
m
a
cr
o
av
er
a
g
es
y
ield
d
if
f
er
en
t
in
ter
p
r
etatio
n
s
:
m
icr
o
av
er
ag
es
d
escr
ib
e
o
v
er
all
p
er
f
o
r
m
a
n
ce
,
wh
ile
m
ac
r
o
av
er
ag
es
eq
u
alize
p
er
f
o
r
m
an
ce
ac
r
o
s
s
class
es,
r
e
v
ea
lin
g
p
o
ten
tial im
b
alan
ce
s
[
7
]
.
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
S
ta
ck
in
g
e
n
s
emb
le
fo
r
mu
ltil
in
g
u
a
l a
s
p
ec
t
-
b
a
s
ed
s
en
timen
t a
n
a
lysi
s
o
f to
u
r
is
m
…
(
B
a
s
w
o
r
o
A
r
d
i P
r
a
mo
n
o
)
4385
I
n
th
e
co
n
tex
t
o
f
m
u
ltil
ab
el
class
if
icatio
n
f
o
r
AB
SA
in
m
u
ltil
in
g
u
al
to
u
r
is
m
r
e
v
iews,
ev
alu
atio
n
m
etr
ics
s
u
ch
as
p
r
ec
is
io
n
,
r
ec
all,
F1
-
s
co
r
e
(
m
icr
o
an
d
m
ac
r
o
)
,
Ham
m
in
g
lo
s
s
,
an
d
s
u
b
s
et
ac
cu
r
ac
y
ar
e
ess
en
tial
f
o
r
ac
cu
r
atel
y
ass
e
s
s
in
g
th
e
m
o
d
el'
s
p
er
f
o
r
m
an
ce
.
Pre
cisi
o
n
m
ea
s
u
r
es
th
e
ac
cu
r
ac
y
o
f
th
e
m
o
d
el'
s
p
o
s
itiv
e
p
r
ed
ictio
n
s
[
1
8
]
.
On
th
e
o
th
er
h
an
d
,
r
ec
all
m
ea
s
u
r
es
th
e
m
o
d
el'
s
ab
ilit
y
to
ca
p
tu
r
e
all
r
elev
a
n
t
in
s
tan
ce
s
[
1
9
]
.
F1
-
s
co
r
es
th
at
co
m
b
in
e
p
r
ec
is
io
n
an
d
r
ec
all
ar
e
p
ar
ticu
lar
l
y
u
s
ef
u
l
in
m
u
ltil
ab
el
task
s
[
2
0
]
.
Ham
m
in
g
lo
s
s
is
th
e
p
r
o
p
o
r
t
io
n
o
f
lab
els
p
r
ed
icted
to
b
e
in
co
r
r
ec
t
[
2
1
]
.
Su
b
s
et
ac
cu
r
ac
y
m
ea
s
u
r
es
th
e
p
r
o
p
o
r
tio
n
o
f
lab
el
p
r
ed
ictio
n
s
th
at
m
atch
th
e
ac
tu
al
lab
el.
T
h
ese
m
etr
ics
h
a
v
e
b
e
en
wid
ely
u
s
ed
in
AB
SA
s
ettin
g
s
an
d
h
av
e
p
r
o
v
en
r
el
ev
an
t
f
o
r
m
ea
s
u
r
in
g
th
e
p
er
f
o
r
m
an
ce
o
f
s
en
tim
en
t
a
n
aly
s
is
m
o
d
els
an
d
th
eir
r
esil
ien
ce
in
h
an
d
lin
g
m
u
ltil
ab
el
d
ata,
esp
ec
ially
in
u
n
b
alan
c
ed
class
if
icatio
n
[
2
2
]
.
Pre
cisi
o
n
m
ea
s
u
r
es
h
o
w
ac
cu
r
ately
th
e
m
o
d
el'
s
p
o
s
itiv
e
p
r
ed
ictio
n
s
ar
e.
T
h
is
ex
p
lain
s
t
h
e
co
r
r
ec
t
p
r
o
p
o
r
tio
n
o
f
th
e
p
r
ed
ictio
n
o
f
th
e
o
v
er
all
p
o
s
itiv
e
p
r
ed
ictio
n
,
as sh
o
wn
in
(
1
)
.
=
+
(
1
)
R
ec
all
m
ea
s
u
r
es
th
e
ex
ten
t
t
o
wh
ich
th
e
m
o
d
el
ca
n
ca
p
t
u
r
e
all
o
f
th
e
ac
tu
al
p
o
s
itiv
e
in
s
tan
ce
s
.
T
h
is
is
im
p
o
r
tan
t to
a
v
o
id
m
is
tak
es in
s
k
ip
p
in
g
lab
els th
at
s
h
o
u
ld
h
a
v
e
b
ee
n
p
r
ed
icted
,
as sh
o
wn
in
(
2
)
.
=
+
(
2
)
F1
-
s
co
r
e
is
a
h
ar
m
o
n
ic
m
ea
n
o
f
p
r
ec
is
io
n
an
d
r
ec
all,
p
r
o
v
id
in
g
a
b
alan
ce
d
ass
ess
m
en
t,
esp
ec
ially
wh
en
lab
el
d
is
tr
ib
u
tio
n
s
ar
e
i
m
b
alan
ce
d
.
T
h
er
e
ar
e
two
ty
p
es:
F1
-
m
icr
o
an
d
F1
-
m
ac
r
o
.
F1
-
m
icr
o
ca
lcu
lates
th
e
to
tal
tr
u
e
p
o
s
itiv
es,
f
alse
p
o
s
itiv
es,
an
d
f
alse
n
eg
ativ
es
in
ag
g
r
e
g
ate
ac
r
o
s
s
all
lab
els.
T
h
e
F1
-
m
ac
r
o
ca
lcu
lates
th
e
F1
f
o
r
ea
ch
lab
el
in
d
iv
id
u
ally
an
d
th
en
a
v
er
ag
es
it.
T
h
is
g
iv
es
eq
u
al
atten
tio
n
to
ea
ch
class
,
in
clu
d
in
g
t
h
e
r
ar
e
o
n
es
,
as sh
o
wn
in
(
3
)
.
1
−
=
2
×
×
+
(
3
)
Ham
m
in
g
l
o
s
s
is
a
m
etr
ic
th
at
m
ea
s
u
r
es
th
e
av
er
ag
e
n
u
m
b
er
o
f
m
is
class
if
icatio
n
s
p
er
lab
el.
I
n
th
e
co
n
tex
t
o
f
m
u
ltil
ab
el
class
if
icatio
n
,
th
e
Ham
m
in
g
lo
s
s
is
th
e
p
r
o
p
o
r
tio
n
o
f
in
co
r
r
ec
tl
y
p
r
e
d
icted
lab
els
(
b
o
th
f
alse
p
o
s
itiv
es
an
d
f
alse
n
eg
ativ
es)
r
elativ
e
t
o
th
e
t
o
tal
n
u
m
b
e
r
o
f
lab
els.
T
h
e
l
o
wer
th
e
Ham
m
in
g
lo
s
s
,
th
e
b
etter
th
e
m
o
d
el'
s
p
er
f
o
r
m
an
ce
.
A
v
alu
e
o
f
0
m
ea
n
s
th
at
all
lab
els ar
e
co
r
r
ec
tly
class
if
ied
,
as sh
o
wn
in
(
4
)
.
l
=
1
×
∑
∑
1
[
≠
ŷ
]
=
1
=
1
(
4
)
W
h
er
e
is
n
u
m
b
er
o
f
s
am
p
les
,
is
n
u
m
b
er
o
f
lab
els p
er
s
am
p
l
e
,
an
d
≠
ŷ
is
a
ctu
al
an
d
p
r
ed
icted
lab
els
f
o
r
th
e
J
-
lab
el
i
n
th
e
I
-
I
s
am
p
l
e
.
T
h
e
ac
cu
r
ac
y
s
u
b
s
et
is
a
m
etr
i
c
th
at
s
h
o
ws
th
e
p
e
r
ce
n
tag
e
o
f
a
s
am
p
le
in
wh
ic
h
th
e
p
r
ed
ict
ed
lab
el
is
ex
ac
tly
th
e
s
am
e
as
th
e
ac
tu
al
lab
el.
T
h
is
m
etr
ic
is
v
er
y
s
tr
ict.
I
f
th
er
e
is
o
n
ly
o
n
e
wr
o
n
g
l
ab
el
in
a
m
u
ltil
ab
el
p
r
ed
ictio
n
,
th
en
all
p
r
ed
ictio
n
s
ar
e
co
n
s
id
er
ed
f
alse
f
o
r
t
h
at
s
am
p
le.
T
h
e
h
ig
h
er
th
e
s
u
b
s
et
ac
cu
r
ac
y
,
t
h
e
m
o
r
e
p
r
ec
is
e
th
e
m
o
d
el'
s
p
r
ed
ictio
n
s
ar
e
in
h
an
d
lin
g
lab
el
c
o
m
b
in
atio
n
s
,
as sh
o
wn
in
(
5
)
.
Sub
s
e
t
a
c
c
ura
c
y
=
1
∑
1
=
Ŷ
=
1
(
5
)
W
h
er
e
Yᵢ
d
an
Ŷᵢ
is
a
ctu
al
an
d
p
r
ed
icted
lab
el
s
ets f
o
r
t
h
e
i th
s
am
p
le
.
2
.
8
.
E
ns
em
ble
t
ra
ini
ng
a
nd
re
-
ev
a
lua
t
io
n
On
ce
th
e
b
aselin
e
m
o
d
els
ar
e
ev
alu
ated
,
th
e
f
o
cu
s
s
h
if
ts
to
tr
ain
in
g
an
d
ev
alu
atin
g
t
h
e
ass
em
b
ly
m
o
d
els:
v
o
tin
g
,
Ad
aBo
o
s
t
,
a
n
d
s
tack
in
g
.
T
h
e
e
n
s
em
b
le
a
p
p
r
o
ac
h
,
p
a
r
ticu
lar
ly
s
tack
in
g
,
aim
s
to
co
m
b
in
e
p
r
ed
ictio
n
s
f
r
o
m
b
asic
m
o
d
el
s
to
im
p
r
o
v
e
g
en
er
aliza
tio
n
ca
p
ab
ilit
ies
an
d
th
e
ac
cu
r
ac
y
o
f
f
in
al
p
r
ed
ictio
n
s
.
E
v
alu
atin
g
th
e
p
er
f
o
r
m
an
ce
o
f
en
s
em
b
le
s
tack
in
g
m
eth
o
d
s
,
s
u
ch
as
v
o
tin
g
,
Ad
aBo
o
s
t,
an
d
s
tack
in
g
,
f
o
r
m
u
ltil
in
g
u
al
AB
SA
in
to
u
r
is
m
r
ev
iews
is
cr
itical
g
iv
en
th
e
co
m
p
le
x
n
atu
r
e
o
f
s
en
tim
en
t
ex
p
r
ess
ed
ac
r
o
s
s
d
if
f
er
en
t
lan
g
u
ag
es
an
d
cu
ltu
r
al
co
n
tex
ts
.
E
n
s
em
b
le
m
et
h
o
d
s
h
av
e
b
ee
n
s
h
o
wn
t
o
im
p
r
o
v
e
class
if
icatio
n
ac
cu
r
ac
y
b
y
lev
er
ag
i
n
g
th
e
c
o
llectiv
e
s
tr
en
g
th
o
f
m
u
ltip
le
m
o
d
els,
th
er
eb
y
r
ed
u
ci
n
g
th
e
li
m
itatio
n
s
ass
o
ciate
d
with
in
d
iv
id
u
al
class
if
ier
s
[
2
3
]
.
T
h
ese
m
eth
o
d
s
ar
e
p
ar
ticu
l
ar
ly
b
en
e
f
icial
in
to
u
r
is
m
,
wh
er
e
s
en
tim
en
t
ab
o
u
t
v
ar
io
u
s
asp
ec
ts
s
u
ch
as
f
ac
ilit
ies,
attr
ac
tio
n
s
,
an
d
ac
ce
s
s
ib
ilit
y
ca
n
v
ar
y
s
ig
n
i
f
ican
tly
ac
r
o
s
s
lan
g
u
ag
es,
m
ak
in
g
th
e
c
h
allen
g
es
m
o
r
e
p
r
o
n
o
u
n
ce
d
[
2
4
]
,
[
2
5
]
.
B
y
in
te
g
r
atin
g
v
a
r
io
u
s
alg
o
r
ith
m
s
,
th
e
en
s
em
b
le
s
tack
in
g
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.
15
,
No
.
5
,
Octo
b
er
2
0
2
6
:
4
3
8
1
-
4
3
9
1
4386
m
eth
o
d
o
f
f
er
s
in
cr
ea
s
ed
r
esis
tan
ce
to
n
o
is
e
an
d
b
etter
g
e
n
er
aliza
tio
n
to
u
n
s
ee
n
d
ata,
wh
ich
is
esp
ec
ially
im
p
o
r
tan
t
wh
en
d
ea
lin
g
wit
h
th
e
u
n
b
alan
c
ed
d
atasets
o
f
ten
en
c
o
u
n
ter
e
d
in
s
en
tim
en
t
an
aly
s
is
[
2
6
]
.
I
n
ad
d
itio
n
,
th
e
ab
ilit
y
o
f
th
e
s
tack
ed
en
s
em
b
le
to
co
m
b
i
n
e
p
r
ed
ictio
n
s
f
r
o
m
d
if
f
er
e
n
t c
lass
if
ie
r
s
p
r
o
v
id
es a
m
o
r
e
n
u
an
ce
d
u
n
d
e
r
s
tan
d
in
g
o
f
th
e
s
en
tim
en
ts
ex
p
r
ess
ed
in
m
u
ltil
in
g
u
al
r
ev
iews,
lead
in
g
to
m
o
r
e
r
eliab
le
in
s
ig
h
ts
f
o
r
s
tak
eh
o
l
d
er
s
in
th
e
to
u
r
is
m
in
d
u
s
tr
y
[
2
7
]
,
[
2
8
]
.
T
h
is
,
in
tu
r
n
,
ca
n
d
r
iv
e
p
r
o
g
r
ess
in
au
to
m
ated
o
p
in
io
n
m
in
in
g
an
d
im
p
r
o
v
e
cu
s
to
m
e
r
s
atis
f
ac
tio
n
th
r
o
u
g
h
tailo
r
ed
r
ec
o
m
m
en
d
atio
n
s
b
ased
o
n
ac
cu
r
ate
s
en
tim
en
t
ass
es
s
m
en
ts
[
2
9
]
.
T
h
e
e
n
s
em
b
le
m
o
d
el
was
th
en
r
e
-
e
v
alu
ated
u
s
in
g
th
e
s
am
e
m
et
r
ic
to
c
o
m
p
ar
e
its
p
er
f
o
r
m
an
ce
im
p
r
o
v
e
m
en
ts
r
e
lativ
e
to
th
e
b
aselin
e
m
o
d
el
an
d
b
etwe
en
f
ello
w
e
n
s
em
b
le
m
o
d
els
[
3
0
]
.
2
.
9
.
B
est
mo
del selec
t
io
n
T
h
e
b
est
m
o
d
els
ar
e
s
elec
ted
b
ased
o
n
th
e
h
ig
h
est
ag
g
r
eg
ate
p
er
f
o
r
m
a
n
ce
ac
r
o
s
s
all
ev
alu
atio
n
m
etr
ics.
I
n
th
is
s
tu
d
y
,
th
e
s
ta
ck
in
g
en
s
em
b
le
s
h
o
wed
th
e
b
est
p
er
f
o
r
m
a
n
ce
an
d
was
p
r
o
p
o
s
ed
as
th
e
m
o
s
t
ac
cu
r
ate
an
d
r
eliab
le
m
o
d
el
f
o
r
AB
SA
o
n
m
u
ltil
in
g
u
al
to
u
r
r
ev
iews
[
1
4
]
,
[
2
9
]
.
T
h
e
s
ec
o
n
d
ar
y
o
b
jectiv
e
is
to
id
en
tify
th
e
b
est
-
p
er
f
o
r
m
in
g
m
o
d
els
f
o
r
to
u
r
is
m
r
ec
o
m
m
e
n
d
atio
n
s
y
s
tem
s
o
r
au
to
m
ated
o
p
in
io
n
p
o
llin
g
to
s
u
p
p
o
r
t
u
s
er
an
d
s
er
v
ice
p
r
o
v
id
er
d
ec
is
io
n
-
m
a
k
in
g
.
T
h
is
r
e
s
ea
r
ch
also
h
elp
s
b
r
id
g
e
th
e
g
ap
in
m
u
ltil
in
g
u
a
l
AB
SA
b
y
em
p
h
asizin
g
k
ey
a
s
p
ec
ts
o
f
to
u
r
is
m
,
s
u
ch
as
am
en
ities
,
attr
ac
tio
n
s
,
an
d
ac
ce
s
s
ib
ilit
y
[
1
]
.
E
th
ical
co
n
s
id
er
atio
n
s
ar
e
also
tak
en
in
to
ac
co
u
n
t,
as
th
e
d
ata
is
p
u
b
lic,
an
o
n
y
m
o
u
s
,
an
d
d
o
es
n
o
t c
o
n
tain
p
er
s
o
n
al
o
r
s
en
s
itiv
e
in
f
o
r
m
atio
n
.
2
.
1
0
.
T
ec
hn
ica
l
im
plem
ent
a
t
io
n a
nd
v
is
ua
liza
t
io
n
T
h
e
en
tire
ex
p
er
im
en
tal
s
tag
e
,
f
r
o
m
d
ata
p
r
ep
r
o
ce
s
s
in
g
to
m
o
d
el
ev
alu
atio
n
,
is
im
p
lem
e
n
ted
u
s
in
g
Py
th
o
n
(
v
e
r
s
io
n
3
.
1
2
.
1
2
)
.
T
h
is
s
tu
d
y
'
s
m
ain
Py
th
o
n
lib
r
ar
ies
in
clu
d
e
Pan
d
as
f
o
r
d
ata
m
an
ip
u
latio
n
,
s
cik
it
-
lear
n
f
o
r
im
p
lem
en
tin
g
m
ac
h
in
e
lear
n
i
n
g
m
o
d
els
an
d
ev
alu
atio
n
m
etr
ics,
an
d
a
st
f
o
r
ea
r
ly
-
s
tag
e
lab
el
d
ata
tr
an
s
f
o
r
m
atio
n
.
All
ex
p
e
r
im
en
ts
wer
e
r
u
n
u
s
in
g
Py
th
o
n
with
th
e
s
cik
it
-
lear
n
lib
r
ar
y
f
o
r
m
o
d
el
tr
ain
in
g
an
d
ev
alu
atio
n
,
as
well
as
f
o
r
th
e
ap
p
licatio
n
o
f
class
if
icatio
n
m
etr
ics.
Descr
ip
tiv
e
s
t
atis
tics
ar
e
u
s
ed
to
s
u
m
m
ar
ize
th
e
m
o
d
el'
s
p
er
f
o
r
m
an
ce
.
Fo
r
g
r
ap
h
ical
v
is
u
aliza
tio
n
o
f
m
o
d
el
r
esu
lts
in
co
m
p
ar
is
o
n
,
wh
ich
s
u
p
p
o
r
ts
m
o
r
e
ef
f
ec
tiv
e
in
ter
p
r
etatio
n
an
d
d
eliv
e
r
y
o
f
d
at
a
-
b
ased
in
f
o
r
m
atio
n
,
th
e
Ma
tp
lo
tlib
an
d
Seab
o
r
n
lib
r
ar
ies ar
e
u
s
ed
.
C
o
m
p
u
te
ex
p
er
im
en
ts
r
u
n
in
a
Go
o
g
le
C
o
l
ab
o
r
ato
r
y
en
v
ir
o
n
m
e
n
t.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
3
.
1
.
Resul
t
s
o
f
ba
s
eline
a
nd
ens
em
ble m
o
del e
v
a
lua
t
io
n
T
h
is
s
tu
d
y
p
r
esen
ts
th
e
r
esu
lt
s
o
f
e
v
alu
atio
n
s
o
f
th
r
ee
b
as
elin
e
m
ac
h
in
e
lear
n
in
g
m
o
d
els,
n
am
ely
lo
g
is
tic
r
eg
r
ess
io
n
,
n
aiv
e
B
ay
es,
an
d
r
an
d
o
m
f
o
r
est
,
as
well
as
th
r
ee
as
s
em
b
ly
m
o
d
e
ls
:
v
o
tin
g
class
if
ier
,
Ad
aBo
o
s
t
class
if
ier
,
an
d
s
tack
in
g
class
if
ier
.
T
h
e
o
p
tim
al
co
n
f
ig
u
r
atio
n
f
o
r
th
e
s
tack
in
g
clas
s
if
ier
u
s
es
lo
g
is
ti
c
r
eg
r
ess
io
n
an
d
r
a
n
d
o
m
f
o
r
es
t
as
th
e
b
ase
m
o
d
el
(
b
ase
lear
n
er
s
)
lev
el
0
,
a
n
d
lo
g
is
tic
r
eg
r
ess
io
n
as
th
e
m
eta
-
lear
n
er
(
lev
el
1
)
,
with
t
h
e
p
ass
th
r
o
u
g
h
=
tr
u
e
o
p
tio
n
.
T
h
e
e
n
tire
m
o
d
el
is
im
p
lem
e
n
ted
u
s
in
g
th
e
Ov
R
c
lass
if
ier
wr
ap
p
er
to
h
a
n
d
le
th
e
m
u
ltil
ab
el
class
if
icatio
n
p
r
o
p
er
ties
o
f
th
is
task
.
Mo
d
el
p
er
f
o
r
m
an
ce
was
ev
alu
ated
u
s
in
g
a
s
et
o
f
s
tan
d
ar
d
m
etr
ics
f
o
r
m
u
ltil
ab
el
clas
s
if
icatio
n
,
in
clu
d
in
g
F1
-
s
co
r
e
(
m
icr
o
a
n
d
m
ac
r
o
av
er
ag
e
d
)
,
p
r
ec
is
i
o
n
(
m
icr
o
av
er
a
g
ed
)
,
r
ec
all
(
m
icr
o
av
e
r
ag
ed
)
,
Ham
m
in
g
lo
s
s
,
an
d
s
u
b
s
et
ac
cu
r
ac
y
.
T
h
ese
m
etr
ics
p
r
o
v
id
e
a
co
m
p
r
eh
e
n
s
iv
e
v
iew
o
f
th
e
m
o
d
el'
s
p
r
ed
ictio
n
ac
cu
r
ac
y
,
g
en
er
aliza
tio
n
ca
p
a
b
ilit
ies,
an
d
er
r
o
r
r
ate.
T
ab
le
1
p
r
esen
ts
th
e
co
m
p
ar
ativ
e
r
esu
lts
o
f
all
th
e
m
o
d
els
test
ed
o
n
th
e
test
d
ataset.
T
h
e
v
alu
es p
r
esen
ted
ar
e
th
e
av
er
a
g
e
p
er
f
o
r
m
an
ce
o
b
tain
e
d
.
T
ab
le
1
.
AB
SA m
o
d
el
ev
alu
at
io
n
r
esu
lts
b
ased
o
n
m
u
ltil
ab
el
m
etr
ics
M
o
d
e
l
F1
-
mi
c
r
o
F1
-
mac
r
o
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
H
a
mm
i
n
g
l
o
ss
S
u
b
s
e
t
a
c
c
u
r
a
c
y
Lo
g
i
s
t
i
c
r
e
g
r
e
ssi
o
n
0
.
7
9
6
0
.
5
7
1
0
.
8
8
1
0
.
7
2
5
0
.
0
8
8
0
.
5
8
8
N
a
i
v
e
B
a
y
e
s
0
.
6
9
4
0
.
4
2
3
0
.
8
1
2
0
.
6
0
6
0
.
1
2
6
0
.
4
5
9
R
a
n
d
o
m f
o
r
e
s
t
0
.
8
3
1
0
.
5
9
0
0
.
8
7
8
0
.
7
8
9
0
.
0
7
6
0
.
6
4
0
V
o
t
i
n
g
0
.
8
00
0
.
5
2
4
0
.
8
8
5
0
.
7
2
9
0
.
0
8
6
0
.
6
0
1
A
d
a
B
o
o
st
0
.
7
9
9
0
.
6
6
6
0
.
8
0
6
0
.
7
9
3
0
.
0
9
4
0
.
5
6
4
S
t
a
c
k
i
n
g
0
.
8
6
3
0
.
7
2
9
0
.
8
8
5
0
.
8
4
1
0
.
0
6
3
0
.
6
8
2
3
.
2
.
M
o
del c
o
m
pa
riso
n v
is
ua
liza
t
io
n
Fig
u
r
e
3
co
m
p
ar
es
th
e
F1
-
m
icr
o
a
n
d
F1
-
m
ac
r
o
s
co
r
es
o
f
all
e
v
alu
ated
b
aselin
e
a
n
d
en
s
em
b
le
m
o
d
els.
T
h
is
g
r
ap
h
v
is
u
ally
co
m
p
ar
es
th
e
F1
-
s
co
r
e
v
alu
es
(
m
icr
o
an
d
m
ac
r
o
)
f
o
r
all
m
o
d
els
tes
ted
.
Fro
m
th
is
v
is
u
aliza
tio
n
,
th
e
s
u
p
e
r
io
r
ity
o
f
th
e
s
tack
in
g
en
s
em
b
le
in
b
o
t
h
F1
-
s
co
r
e
m
etr
ics
b
ec
o
m
es
v
is
ib
le
co
m
p
ar
ed
to
th
e
o
th
er
m
o
d
els.
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
S
ta
ck
in
g
e
n
s
emb
le
fo
r
mu
ltil
in
g
u
a
l a
s
p
ec
t
-
b
a
s
ed
s
en
timen
t a
n
a
lysi
s
o
f to
u
r
is
m
…
(
B
a
s
w
o
r
o
A
r
d
i P
r
a
mo
n
o
)
4387
Fig
u
r
e
3
.
C
o
m
p
a
r
is
o
n
o
f
F1
-
m
icr
o
an
d
F1
-
m
ac
r
o
ac
r
o
s
s
th
e
ev
alu
ated
m
o
d
els
3
.
3
.
Descript
io
n o
f
re
s
ults per
m
o
del
A
m
o
r
e
d
etailed
a
n
aly
s
is
o
f
th
e
r
esu
lts
in
T
ab
le
1
f
o
r
ea
ch
m
o
d
el
is
as f
o
llo
ws.
3
.
3
.
1
.
L
o
g
is
t
ic
re
g
re
s
s
io
n
L
o
g
is
tic
r
eg
r
ess
io
n
s
h
o
ws
s
ta
b
le
b
aselin
e
p
er
f
o
r
m
an
ce
f
o
r
m
u
ltil
in
g
u
al
m
u
ltil
ab
el
AB
SA
,
p
r
o
v
i
d
in
g
a
s
tr
o
n
g
lin
ea
r
r
ef
e
r
en
ce
u
n
d
e
r
T
F
-
I
DF
f
ea
t
u
r
es.
T
h
e
m
o
d
el
ac
h
iev
es
a
n
F1
-
m
icr
o
o
f
0
.
7
9
6
an
d
a
n
F1
-
m
ac
r
o
o
f
0
.
5
7
1
,
with
p
r
ec
is
io
n
0
.
8
8
1
an
d
r
ec
all
0
.
7
2
5
.
W
h
ile
th
ese
r
esu
lts
in
d
icate
r
eliab
le
o
v
er
all
d
etec
tio
n
o
f
co
m
m
o
n
lab
els,
th
e
m
o
d
er
ate
m
ac
r
o
-
F1
s
u
g
g
ests
th
at
p
er
f
o
r
m
an
ce
is
less
u
n
if
o
r
m
ac
r
o
s
s
all
asp
ec
t
s
en
tim
en
t
lab
els,
wh
ich
is
ty
p
ical
in
m
u
ltil
ab
el
d
atasets
with
lab
el
im
b
alan
ce
.
T
h
e
Ham
m
in
g
l
o
s
s
o
f
0
.
0
8
8
a
n
d
s
u
b
s
et
ac
cu
r
ac
y
o
f
0
.
5
8
8
f
u
r
th
e
r
s
h
o
w
th
at
ex
ac
t
-
m
atch
p
r
ed
ictio
n
s
r
em
ain
c
h
allen
g
in
g
wh
e
n
m
u
ltip
le
lab
els
m
u
s
t
b
e
jo
in
tly
p
r
ed
icted
.
3
.
3
.
2
.
Na
iv
e
B
a
y
es
Mu
ltin
o
m
ial
n
aiv
e
B
ay
es
p
er
f
o
r
m
s
wo
r
s
t
am
o
n
g
th
e
e
v
al
u
ated
m
o
d
els,
r
ef
lectin
g
its
s
im
p
lify
in
g
in
d
ep
en
d
en
ce
ass
u
m
p
tio
n
s
in
co
m
p
lex
m
u
ltil
in
g
u
al
r
ev
iew
p
atter
n
s
.
I
t
r
ec
o
r
d
s
an
F1
-
m
ic
r
o
o
f
0
.
6
9
4
a
n
d
a
n
F1
-
m
ac
r
o
o
f
0
.
4
2
3
,
with
p
r
ec
i
s
io
n
0
.
8
1
2
a
n
d
r
ec
all
0
.
6
0
6
.
T
h
e
r
elativ
ely
l
o
w
m
ac
r
o
-
F1
in
d
icate
s
th
at
s
ev
er
al
lab
els,
p
ar
ticu
lar
ly
less
f
r
eq
u
en
t
asp
ec
t
-
s
en
tim
en
t
co
m
b
in
ati
o
n
s
,
ar
e
n
o
t
co
n
s
is
ten
tly
ca
p
tu
r
ed
.
T
h
is
lim
itatio
n
is
also
r
ef
lecte
d
in
th
e
h
ig
h
er
Ham
m
in
g
l
o
s
s
(
0
.
1
2
6
)
an
d
th
e
lo
west
s
u
b
s
et
ac
cu
r
ac
y
(
0
.
4
5
9
)
,
s
h
o
win
g
d
if
f
icu
lty
in
p
r
o
d
u
cin
g
f
u
lly
c
o
r
r
ec
t m
u
ltil
ab
el
o
u
tp
u
ts
.
3
.
3
.
3
.
Ra
nd
o
m
f
o
re
s
t
R
an
d
o
m
f
o
r
est
p
r
o
v
id
es
a
s
tr
o
n
g
n
o
n
-
lin
ea
r
b
aselin
e
an
d
p
er
f
o
r
m
s
co
m
p
etitiv
ely
ac
r
o
s
s
th
e
ev
alu
ated
m
etr
ics.
T
h
e
m
o
d
el
ac
h
iev
es
an
F1
-
m
icr
o
o
f
0
.
8
3
1
an
d
a
n
F1
-
m
ac
r
o
o
f
0
.
5
9
0
,
with
p
r
ec
is
io
n
0
.
8
7
8
an
d
r
ec
all
0
.
7
8
9
.
C
o
m
p
ar
ed
t
o
lin
ea
r
b
aselin
es,
r
an
d
o
m
f
o
r
ests
b
en
ef
it
f
r
o
m
m
o
d
elin
g
f
ea
tu
r
e
in
ter
ac
tio
n
s
,
wh
ich
ca
n
b
etter
ca
p
tu
r
e
h
ete
r
o
g
en
e
o
u
s
wo
r
d
i
n
g
in
t
o
u
r
is
m
r
ev
iews.
I
ts
Ham
m
in
g
lo
s
s
d
ec
r
ea
s
es
to
0
.
0
7
6
an
d
s
u
b
s
e
t
ac
cu
r
a
c
y
i
n
cr
ea
s
es
to
0
.
6
4
0
,
in
d
i
ca
t
in
g
i
m
p
r
o
v
e
d
m
u
ltil
a
b
el
c
o
n
s
is
t
en
c
y
an
d
f
ew
er
p
e
r
-
l
a
b
el
e
r
r
o
r
s
.
3
.
3
.
4
.
Vo
t
ing
cla
s
s
if
ier
T
h
e
v
o
tin
g
e
n
s
em
b
le
p
r
o
d
u
ce
s
in
ter
m
ed
iate
p
er
f
o
r
m
a
n
c
e
b
y
co
m
b
in
i
n
g
co
m
p
lem
en
tar
y
b
ase
lear
n
er
s
th
r
o
u
g
h
p
r
o
b
a
b
ilit
y
a
g
g
r
eg
atio
n
.
I
t
ac
h
ie
v
es
an
F1
-
m
icr
o
o
f
0
.
8
0
0
a
n
d
an
F1
-
m
a
cr
o
o
f
0
.
5
2
4
,
with
p
r
ec
is
io
n
0
.
8
8
5
an
d
r
ec
all
0
.
7
2
9
.
Alth
o
u
g
h
th
e
en
s
em
b
le
s
t
ab
ilizes
p
r
ed
ictio
n
s
r
elativ
e
to
s
in
g
le
lear
n
e
r
s
,
th
e
m
ac
r
o
-
F1
r
e
m
ain
s
lo
wer
th
an
s
tr
o
n
g
er
e
n
s
em
b
les,
in
d
icatin
g
th
at
m
in
o
r
ity
lab
els
s
till
r
ec
eiv
e
lim
ited
g
ain
s
.
T
h
is
p
atter
n
is
co
n
s
is
ten
t
w
i
th
its
Ham
m
in
g
l
o
s
s
(
0
.
0
8
6
)
an
d
s
u
b
s
et
ac
cu
r
ac
y
(
0
.
6
0
1
)
,
wh
ich
im
p
r
o
v
e
m
o
d
estl
y
b
u
t
d
o
n
o
t m
atch
th
e
b
est
-
p
er
f
o
r
m
in
g
m
o
d
el.
3
.
3
.
5
.
Ada
B
o
o
s
t
cla
s
s
if
ier
Ad
aBo
o
s
t
im
p
r
o
v
es
p
er
f
o
r
m
a
n
ce
b
y
s
eq
u
en
tially
f
o
cu
s
in
g
o
n
m
is
class
if
ied
in
s
tan
ce
s
,
wh
ich
ca
n
b
e
b
en
ef
icial
wh
en
s
o
m
e
asp
ec
ts
o
f
s
en
tim
en
t
lab
els
ar
e
h
ar
d
e
r
to
d
etec
t.
T
h
e
m
o
d
el
ac
h
ie
v
es
an
F1
-
m
icr
o
o
f
0
.
7
9
9
an
d
a
n
o
tab
ly
h
ig
h
er
F1
-
m
ac
r
o
o
f
0
.
6
6
6
,
with
p
r
ec
is
io
n
0
.
8
0
6
an
d
r
ec
all
0
.
7
9
3
.
T
h
e
h
ig
h
er
m
ac
r
o
-
F1
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.
15
,
No
.
5
,
Octo
b
er
2
0
2
6
:
4
3
8
1
-
4
3
9
1
4388
s
u
g
g
ests
th
at
Ad
aBo
o
s
t
y
iel
d
s
m
o
r
e
b
alan
ce
d
lab
el
-
wis
e
p
er
f
o
r
m
an
ce
th
an
s
ev
e
r
al
b
aselin
es,
p
o
ten
tially
im
p
r
o
v
in
g
wea
k
e
r
lab
els.
Ho
wev
er
,
its
Ham
m
in
g
lo
s
s
(
0
.
0
9
4
)
a
n
d
s
u
b
s
et
ac
cu
r
ac
y
(
0
.
5
6
4
)
in
d
icate
th
at
e
x
ac
t
m
u
ltil
ab
el
m
atch
es r
em
ain
d
i
f
f
icu
lt a
n
d
th
at
g
ain
s
ar
e
n
o
t u
n
if
o
r
m
ac
r
o
s
s
all
ev
alu
atio
n
cr
it
er
ia.
3
.
3
.
6
.
Sta
c
k
ing
cla
s
s
if
ier
T
h
e
s
tack
in
g
class
if
ier
co
n
s
i
s
ten
tly
d
em
o
n
s
tr
ates
th
e
b
est
o
v
er
all
p
er
f
o
r
m
a
n
ce
ac
r
o
s
s
th
e
m
ajo
r
ev
alu
atio
n
m
etr
ics.
I
t
ac
h
iev
e
s
th
e
h
ig
h
est
F1
-
m
icr
o
(
0
.
8
6
2
6
)
an
d
F1
-
m
ac
r
o
(
0
.
7
2
8
5
)
,
in
d
icatin
g
b
o
th
s
tr
o
n
g
o
v
er
all
ac
cu
r
ac
y
an
d
b
alan
ce
d
lab
el
-
wis
e
p
er
f
o
r
m
an
ce
.
I
n
ad
d
itio
n
,
it
p
r
o
d
u
ce
s
th
e
lo
west
Ham
m
in
g
lo
s
s
(
0
.
0
6
3
3
)
a
n
d
th
e
h
ig
h
est
s
u
b
s
et
ac
cu
r
ac
y
(
0
.
6
8
2
5
)
,
s
h
o
win
g
f
ewe
r
lab
el
er
r
o
r
s
an
d
b
etter
ex
ac
t
-
m
atc
h
p
er
f
o
r
m
an
ce
.
T
h
ese
r
esu
lts
s
u
g
g
est
th
at
m
eta
-
lear
n
in
g
ef
f
ec
tiv
ely
co
m
b
in
es
co
m
p
lem
e
n
tar
y
s
tr
en
g
th
s
o
f
th
e
b
ase
lear
n
er
s
(
lo
g
is
tic
r
eg
r
ess
io
n
an
d
r
an
d
o
m
f
o
r
est)
th
r
o
u
g
h
a
lo
g
is
tic
r
eg
r
ess
io
n
m
eta
-
lear
n
er
.
3.
4
.
Dis
cus
s
io
n
T
h
e
r
esu
lts
s
h
o
wed
th
at
th
e
s
tack
in
g
en
s
em
b
le
m
o
d
el
c
o
n
s
is
ten
tly
o
u
tp
er
f
o
r
m
ed
class
ica
l
m
ac
h
in
e
lear
n
in
g
m
o
d
els
an
d
o
th
e
r
en
s
em
b
le
m
eth
o
d
s
in
m
u
ltil
ab
el
class
if
icatio
n
o
f
m
u
ltil
in
g
u
al
to
u
r
r
ev
iews.
T
h
is
is
r
ef
lecte
d
in
th
e
F1
-
s
co
r
e
m
ic
r
o
v
alu
e
o
f
0
.
8
6
2
6
an
d
t
h
e
h
ig
h
er
F1
-
s
co
r
e
m
ac
r
o
o
f
0
.
7
2
8
5
,
as
well
as
th
e
lo
west
Ham
m
in
g
lo
s
s
o
f
0
.
0
6
3
3
an
d
th
e
h
ig
h
est
s
u
b
s
et
o
f
ac
cu
r
ac
y
r
ea
ch
i
n
g
0
.
6
8
2
5
.
I
n
co
m
p
ar
is
o
n
,
th
e
s
tr
o
n
g
est
b
aselin
e
m
o
d
el,
r
an
d
o
m
f
o
r
est
,
ac
h
iev
e
d
a
m
icr
o
F1
-
s
co
r
e
o
f
0
.
8
3
1
3
an
d
a
m
ac
r
o
-
F1
-
s
co
r
e
o
f
0
.
5
8
9
8
,
with
a
H
am
m
in
g
lo
s
s
o
f
0
.
0
7
5
6
an
d
a
s
u
b
s
et
ac
cu
r
ac
y
o
f
0
.
6
4
0
4
.
Oth
er
en
s
em
b
l
e
m
o
d
els,
s
u
ch
as
Ad
aBo
o
s
t,
ac
h
iev
e
m
icr
o
F1
-
s
co
r
es
o
f
0
.
7
9
9
1
an
d
m
ac
r
o
F
1
-
s
co
r
es
o
f
0
.
6
6
5
5
,
w
h
ile
v
o
ti
n
g
ac
h
iev
es
m
icr
o
F1
-
s
co
r
es
o
f
0
.
7
9
9
9
a
n
d
m
ac
r
o
F1
-
s
co
r
es
o
f
0
.
5
2
4
1
.
T
h
ese
f
in
d
in
g
s
r
ei
n
f
o
r
ce
th
e
p
r
em
is
e
th
at
th
e
in
teg
r
atio
n
o
f
m
u
ltip
le
b
asic
m
o
d
els
(
in
th
is
ca
s
e,
lo
g
is
tic
r
eg
r
ess
io
n
an
d
r
an
d
o
m
f
o
r
est
as
b
ase
le
ar
n
er
s
,
with
lo
g
is
tic
r
eg
r
ess
io
n
as
th
e
m
eta
-
lear
n
er
)
t
h
r
o
u
g
h
a
s
tack
in
g
ap
p
r
o
ac
h
ca
n
im
p
r
o
v
e
th
e
g
e
n
er
aliza
tio
n
an
d
ac
cu
r
ac
y
o
f
p
r
ed
ictio
n
s
,
esp
ec
ially
in
t
h
e
c
o
n
tex
t o
f
co
m
p
le
x
an
d
d
iv
er
s
e
d
ata
s
u
ch
as tr
av
el
r
e
v
iews.
I
n
th
is
s
tu
d
y
,
th
e
r
esu
lts
s
h
o
wed
th
at
th
e
s
tack
in
g
en
s
em
b
le
m
eth
o
d
s
ig
n
if
ican
tly
o
u
t
p
er
f
o
r
m
ed
class
ical
m
ac
h
in
e
lear
n
in
g
m
o
d
els
s
u
ch
as
lo
g
is
tic
r
eg
r
ess
io
n
(
F1
-
m
icr
o
:
0
.
7
9
5
8
)
,
n
aiv
e
B
ay
es
(
F1
-
m
icr
o
:
0
.
6
9
4
2
)
,
an
d
r
an
d
o
m
f
o
r
est
(
F1
-
m
icr
o
:
0
.
8
3
1
3
)
,
as
well
as
o
t
h
er
en
s
em
b
le
m
eth
o
d
s
s
u
ch
as
v
o
tin
g
(
F1
-
m
icr
o
:
0
.
7
9
9
9
;
Ham
m
in
g
l
o
s
s
:
0
.
0
8
6
1
;
s
u
b
s
et
ac
cu
r
ac
y
:
0
.
6
0
0
7
)
a
n
d
Ad
aBo
o
s
t
(
F1
-
m
icr
o
:
0
.
7
9
9
1
;
Ham
m
in
g
l
o
s
s
:
0
.
0
9
4
1
;
s
u
b
s
et
ac
cu
r
ac
y
:
0
.
5
6
4
0
)
in
v
ar
io
u
s
k
ey
ev
al
u
atio
n
m
etr
ics.
Sp
ec
if
ically
,
th
e
s
tack
in
g
en
s
em
b
le
n
o
t
o
n
ly
ac
h
ie
v
es
th
e
h
ig
h
est
F
1
-
s
co
r
e
(
m
icr
o
:
0
.
8
6
2
6
,
m
ac
r
o
:
0
.
7
2
8
5
)
b
u
t
also
t
h
e
lo
west
Ham
m
in
g
l
o
s
s
(
0
.
0
6
3
3
)
an
d
th
e
h
i
g
h
est
s
u
b
s
et
ac
cu
r
ac
y
(
0
.
6
8
2
5
)
.
T
h
is
s
u
p
er
io
r
p
er
f
o
r
m
an
ce
o
f
t
h
e
s
tack
in
g
en
s
em
b
l
e
d
em
o
n
s
tr
ates
its
ab
ilit
y
to
co
m
b
in
e
p
r
e
d
ictio
n
s
f
r
o
m
m
u
ltip
le
class
if
ier
s
,
wh
ich
is
p
ar
tic
u
lar
ly
b
en
ef
icial
in
th
e
m
u
ltil
in
g
u
al
lan
d
s
ca
p
e
o
f
to
u
r
is
m
r
ev
iews,
wh
er
e
s
en
t
im
en
t
ex
p
r
ess
io
n
s
ca
n
v
ar
y
wid
ely
[
1
6
]
.
T
h
ese
f
in
d
in
g
s
ar
e
in
lin
e
with
p
r
ev
io
u
s
liter
atu
r
e
[
1
1
]
,
[
1
2
]
,
wh
ich
h
ig
h
lig
h
t
th
e
e
f
f
ec
tiv
en
ess
o
f
en
s
em
b
le
tech
n
iq
u
es
in
s
en
tim
e
n
t
class
if
icatio
n
task
s
,
an
d
s
u
p
p
o
r
t
t
h
e
f
in
d
i
n
g
s
o
f
[
3
]
,
[
8
]
r
e
g
ar
d
in
g
th
e
b
en
ef
its
o
f
s
tack
in
g
f
o
r
a
n
aly
zin
g
d
ata
ac
r
o
s
s
d
ialec
ts
an
d
lan
g
u
ag
es.
B
y
ex
ten
d
in
g
th
ese
r
esu
lts
in
to
th
e
co
n
tex
t
o
f
m
u
ltil
in
g
u
al
AB
SA,
th
e
im
p
licatio
n
s
o
f
u
s
i
n
g
s
tack
in
g
en
s
em
b
l
e
b
ec
o
m
e
s
ig
n
if
ican
t.
I
ts
ab
ilit
y
to
ac
h
iev
e
h
ig
h
er
ac
cu
r
ac
y
o
f
asp
ec
t
an
d
s
en
t
im
en
t
p
r
e
d
ictio
n
s
(
ev
id
en
ce
d
b
y
a
m
icr
o
F1
-
s
co
r
e
o
f
0
.
8
6
2
6
)
ca
n
im
p
r
o
v
e
th
e
ac
cu
r
ac
y
an
d
r
eliab
ilit
y
o
f
s
en
tim
en
t
an
aly
s
is
o
n
to
u
r
is
m
-
r
elate
d
u
s
er
-
g
en
er
ate
d
co
n
ten
t.
T
h
is
,
in
tu
r
n
,
s
u
p
p
o
r
ts
b
etter
d
ec
is
io
n
-
m
a
k
in
g
f
o
r
in
d
u
s
tr
y
s
tak
eh
o
ld
er
s
[
3
1
]
,
[
3
2
]
.
T
h
is
,
i
n
tu
r
n
,
s
u
p
p
o
r
ts
b
etter
d
ec
is
io
n
-
m
ak
in
g
f
o
r
in
d
u
s
tr
y
s
tak
eh
o
ld
er
s
.
Fo
r
ex
am
p
le,
it
ca
n
h
elp
id
en
tif
y
a
d
esti
n
atio
n
'
s
s
p
ec
if
ic
s
tr
en
g
th
s
an
d
wea
k
n
ess
es
b
y
an
aly
zi
n
g
t
h
e
asp
ec
ts
m
o
s
t
f
r
eq
u
e
n
tly
d
is
cu
s
s
ed
,
alo
n
g
with
th
eir
s
en
tim
en
t p
o
lar
ities
.
Ho
wev
er
,
it
s
h
o
u
ld
b
e
n
o
ted
t
h
at
wh
ile
th
e
s
tack
in
g
m
o
d
el
p
er
f
o
r
m
s
well
with
th
e
T
F
-
I
D
F
f
ea
tu
r
e,
r
u
le
-
b
ased
lab
ellin
g
ap
p
r
o
ac
h
es
u
s
ed
in
th
e
ea
r
ly
s
tag
es
o
f
d
ata
c
o
llectio
n
o
r
as
a
c
o
m
p
ar
ato
r
s
till
f
ac
e
ch
allen
g
es,
in
clu
d
in
g
p
o
ten
tia
l
in
ac
cu
r
ac
ies
in
s
en
tim
en
t
class
if
icatio
n
d
u
e
to
co
n
tex
tu
al
lan
g
u
ag
e
n
u
an
ce
s
.
Fu
tu
r
e
r
esear
ch
m
a
y
ex
p
lo
r
e
in
teg
r
atin
g
tr
an
s
f
o
r
m
er
-
b
ased
m
o
d
els
s
u
ch
as
b
id
i
r
ec
tio
n
al
en
co
d
er
r
ep
r
esen
tatio
n
s
f
r
o
m
tr
a
n
s
f
o
r
m
er
s
(
B
E
R
T
)
o
r
cr
o
s
s
-
lin
g
u
al
lan
g
u
ag
e
m
o
d
el
b
ased
R
o
B
E
R
T
a
(
XL
M
-
R
)
with
th
is
p
r
o
v
en
s
tack
in
g
ar
ch
itect
u
r
e.
T
r
a
n
s
f
o
r
m
er
m
o
d
els,
wh
ich
h
av
e
d
e
m
o
n
s
tr
ated
h
i
g
h
er
ac
cu
r
ac
y
th
r
o
u
g
h
co
n
tex
tu
al
u
n
d
er
s
tan
d
in
g
in
s
en
tim
en
t
an
aly
s
is
task
s
[
2
9
]
,
[
3
3
]
,
p
o
ten
tially
f
u
r
th
er
im
p
r
o
v
e
p
er
f
o
r
m
an
ce
,
esp
ec
ially
wh
en
co
m
b
i
n
ed
w
ith
th
e
g
en
er
aliza
tio
n
ca
p
ab
il
ities
o
f
s
tack
in
g
.
I
n
ad
d
itio
n
,
th
e
im
p
o
r
ta
n
ce
o
f
m
u
ltil
ab
el
ev
alu
atio
n
m
et
r
ics
s
u
ch
as
F1
-
s
co
r
e
p
er
lab
el
(
wh
ich
ca
n
also
b
e
ca
lc
u
lated
f
r
o
m
y
o
u
r
m
o
d
el,
e.
g
.
,
f
o
r
F1
_
ac
ce
s
s
ib
ilit
y
_
p
o
s
itiv
e,
F1
_
am
en
ities
_
n
eg
ativ
e)
,
Ham
m
in
g
l
o
s
s
,
an
d
s
u
b
s
et
ac
cu
r
ac
y
is
b
ec
o
m
in
g
in
cr
ea
s
in
g
ly
clea
r
.
T
h
e
m
etr
i
c
is
im
p
o
r
tan
t
f
o
r
co
m
p
r
eh
en
s
iv
ely
an
aly
zin
g
u
s
er
o
p
in
io
n
s
in
a
m
u
ltil
in
g
u
al
AB
SA
s
y
s
tem
,
n
o
t
o
n
ly
s
u
p
p
o
r
tin
g
an
in
-
d
ep
th
u
n
d
er
s
ta
n
d
in
g
o
f
s
en
tim
en
t
r
elate
d
to
v
ar
io
u
s
asp
ec
ts
o
f
to
u
r
is
m
b
u
t a
ls
o
p
r
o
v
i
d
in
g
im
p
o
r
tan
t in
s
ig
h
ts
in
to
u
s
er
p
r
e
f
e
r
en
ce
s
an
d
e
x
p
er
ien
ce
[
3
0
]
,
[
3
4
]
.
T
h
u
s
,
th
is
s
tu
d
y
n
o
t
o
n
ly
f
i
lls
a
m
eth
o
d
o
lo
g
ical
g
ap
in
th
e
m
u
ltil
in
g
u
al
AB
SA
lite
r
atu
r
e
b
y
d
em
o
n
s
tr
atin
g
th
e
s
u
p
er
i
o
r
ity
o
f
th
e
s
tack
in
g
en
s
em
b
le
o
n
th
e
m
u
ltil
in
g
u
al
to
u
r
is
m
r
ev
iew
d
ataset
in
I
n
d
o
n
esia,
b
u
t
also
p
r
o
v
id
es
a
n
im
p
o
r
tan
t
f
o
o
th
o
l
d
f
o
r
f
u
r
th
er
ex
p
lo
r
atio
n
.
T
h
e
s
u
cc
ess
o
f
s
tack
in
g
en
s
em
b
le
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
S
ta
ck
in
g
e
n
s
emb
le
fo
r
mu
ltil
in
g
u
a
l a
s
p
ec
t
-
b
a
s
ed
s
en
timen
t a
n
a
lysi
s
o
f to
u
r
is
m
…
(
B
a
s
w
o
r
o
A
r
d
i P
r
a
mo
n
o
)
4389
with
s
p
ec
if
ic
co
n
f
ig
u
r
atio
n
s
(
l
o
g
is
tic
r
eg
r
ess
io
n
an
d
r
an
d
o
m
f
o
r
est
as
b
ase
lear
n
er
s
,
a
n
d
lo
g
is
tic
r
eg
r
ess
io
n
as
m
eta
-
lear
n
er
s
)
in
ac
h
iev
i
n
g
s
u
b
s
et
ac
cu
r
ac
y
o
f
0
.
6
8
2
5
(
m
ea
n
in
g
th
at
6
8
.
2
5
%
o
f
r
ev
iews
ar
e
p
er
f
ec
tl
y
class
if
ied
f
o
r
all
s
en
tim
en
t
lab
els)
an
d
Ham
m
in
g
l
o
s
s
0
.
0
6
3
3
(
av
e
r
ag
e
er
r
o
r
p
e
r
lab
el
is
v
er
y
lo
w)
b
ec
am
e
b
en
ch
m
ar
k
s
f
o
r
t
h
e
d
ev
elo
p
m
en
t
o
f
a
n
o
p
in
io
n
an
al
y
s
is
an
d
d
esti
n
atio
n
r
ec
o
m
m
en
d
at
io
n
s
y
s
tem
b
ased
o
n
u
s
er
r
ev
iew
d
ata
m
o
r
e
ad
v
an
c
ed
an
d
ac
c
u
r
ate.
4.
CO
NCLU
SI
O
N
T
h
is
s
tu
d
y
b
en
c
h
m
ar
k
e
d
b
a
s
elin
e
an
d
en
s
em
b
le
m
ac
h
in
e
lear
n
in
g
m
eth
o
d
s
f
o
r
m
u
ltil
in
g
u
al
m
u
ltil
ab
el
AB
SA
o
n
to
u
r
is
m
r
ev
iews
co
v
er
in
g
am
en
iti
es,
attr
ac
tio
n
s
,
an
d
ac
ce
s
s
ib
ilit
y
.
Usi
n
g
T
F
-
I
DF
f
ea
tu
r
es
an
d
a
n
Ov
R
m
u
ltil
ab
el
f
r
am
ewo
r
k
,
th
e
r
esu
lts
s
h
o
w
th
at
s
tack
in
g
p
r
o
v
id
es
th
e
s
tr
o
n
g
est
an
d
m
o
s
t
co
n
s
is
ten
t
p
er
f
o
r
m
an
ce
ac
r
o
s
s
F1
-
m
icr
o
,
F1
-
m
ac
r
o
,
Ham
m
in
g
lo
s
s
,
an
d
s
u
b
s
et
ac
cu
r
ac
y
,
o
u
tp
er
f
o
r
m
i
n
g
lo
g
is
tic
r
eg
r
ess
io
n
,
n
aiv
e
B
ay
es,
r
an
d
o
m
f
o
r
est,
v
o
tin
g
,
an
d
Ad
aBo
o
s
t.
T
h
e
f
in
d
i
n
g
s
in
d
icate
th
at
m
eta
-
lear
n
in
g
e
f
f
ec
tiv
ely
co
m
b
in
es
co
m
p
lem
e
n
tar
y
m
o
d
e
l
s
tr
en
g
th
s
an
d
y
ield
s
b
etter
lab
el
-
wis
e
b
alan
ce
,
wh
ich
is
cr
u
cial
f
o
r
e
x
tr
ac
tin
g
r
eliab
le
asp
ec
t
-
s
en
tim
en
t
s
ig
n
als
f
r
o
m
b
ilin
g
u
al
to
u
r
is
m
d
at
a.
I
n
p
r
ac
tice,
th
ese
s
ig
n
als
ca
n
h
elp
to
u
r
is
m
s
tak
eh
o
ld
er
s
p
r
io
r
itize
s
er
v
ice
im
p
r
o
v
em
en
ts
alig
n
ed
with
u
s
er
f
e
ed
b
ac
k
ac
r
o
s
s
th
e
th
r
ee
asp
ec
t
ca
teg
o
r
ies.
Fu
tu
r
e
wo
r
k
s
h
o
u
l
d
v
alid
ate
th
e
ap
p
r
o
ac
h
u
n
d
e
r
ad
d
itio
n
al
r
o
b
u
s
tn
ess
s
ettin
g
s
(
e.
g
.
,
cr
o
s
s
-
v
alid
atio
n
an
d
e
x
ter
n
al
test
d
atasets
)
an
d
co
m
p
ar
e
it
ag
ain
s
t
m
u
ltil
in
g
u
a
l
tr
an
s
f
o
r
m
er
-
b
ased
en
co
d
er
s
to
ass
ess
g
en
er
aliza
tio
n
b
ey
o
n
d
T
F
-
I
DF r
ep
r
esen
tat
io
n
s
.
ACK
NO
WL
E
DG
M
E
N
T
S
T
h
e
au
th
o
r
s
th
an
k
Un
iv
er
s
i
tas
Sem
ar
an
g
an
d
Un
iv
er
s
i
tas
Dip
o
n
eg
o
r
o
f
o
r
th
e
in
s
titu
tio
n
al,
co
m
p
u
tatio
n
al,
an
d
ac
a
d
em
ic
s
u
p
p
o
r
t p
r
o
v
i
d
ed
d
u
r
in
g
th
is
r
esear
ch
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
is
r
esear
ch
r
ec
eiv
ed
in
s
titu
tio
n
al
f
u
n
d
in
g
f
r
o
m
Un
iv
er
s
ita
s
Sem
ar
an
g
.
T
h
e
f
u
n
d
in
g
was
p
r
o
v
id
ed
with
o
u
t a
s
p
ec
if
ic
g
r
a
n
t o
r
c
o
n
tr
ac
t n
u
m
b
e
r
.
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
.
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
B
aswo
r
o
Ar
d
i Pr
am
o
n
o
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
R
ah
m
at
Ger
n
o
wo
✓
✓
✓
✓
✓
✓
✓
Ag
h
u
s
So
f
wan
✓
✓
✓
✓
✓
✓
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
T
h
e
au
th
o
r
s
d
ec
lar
e
n
o
co
n
f
lict o
f
in
ter
est.
I
NF
O
RM
E
D
CO
NS
E
N
T
T
h
is
s
ec
tio
n
n
o
t
ap
p
licab
le.
T
h
is
s
tu
d
y
an
aly
s
es
p
u
b
licly
av
ailab
le
o
n
lin
e
r
ev
iews.
No
h
u
m
a
n
s
u
b
jects we
r
e
r
ec
r
u
ited
,
an
d
n
o
p
er
s
o
n
ally
id
en
tifia
b
le
in
f
o
r
m
atio
n
was c
o
llected
,
s
to
r
ed
,
o
r
r
ep
o
r
ted
.
E
T
H
I
CAL AP
P
RO
V
AL
T
h
is
s
ec
tio
n
n
o
t
ap
p
licab
le.
T
h
e
s
tu
d
y
is
b
ased
o
n
p
u
b
li
cly
av
ailab
le
o
n
lin
e
r
ev
iews.
All
u
s
er
id
en
tifie
r
s
(
e.
g
.
,
u
s
er
n
am
es
an
d
p
r
o
f
ile
lin
k
s
)
wer
e
r
em
o
v
e
d
d
u
r
in
g
p
r
e
p
r
o
ce
s
s
in
g
,
an
d
o
n
ly
an
o
n
y
m
ize
d
tex
t
d
ata
wer
e
an
aly
ze
d
.
T
h
er
ef
o
r
e
,
in
s
titu
tio
n
al
eth
ical
ap
p
r
o
v
al
was n
o
t r
eq
u
ir
e
d
.
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.
15
,
No
.
5
,
Octo
b
er
2
0
2
6
:
4
3
8
1
-
4
3
9
1
4390
DATA AV
AI
L
AB
I
L
I
T
Y
T
h
e
d
ataset
an
aly
ze
d
in
th
is
s
tu
d
y
is
p
u
b
licly
av
ailab
le
o
n
Z
e
n
o
d
o
at
h
ttp
s
://d
o
i.o
r
g
/1
0
.
5
2
8
1
/zen
o
d
o
.
1
5
5
1
7
7
2
7
.
RE
F
E
R
E
NC
E
S
[
1
]
Z.
A
.
-
M
o
u
d
,
H
.
V
.
-
N
e
j
a
d
,
a
n
d
J.
S
a
d
r
i
,
“
T
o
u
r
i
sm
r
e
c
o
mm
e
n
d
a
t
i
o
n
s
y
st
e
m
b
a
se
d
o
n
se
ma
n
t
i
c
c
l
u
st
e
r
i
n
g
a
n
d
s
e
n
t
i
m
e
n
t
a
n
a
l
y
si
s
,
”
Ex
p
e
rt
S
y
st
e
m
s w
i
t
h
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
1
6
7
,
A
p
r
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
sw
a
.
2
0
2
0
.
1
1
4
3
2
4
.
[
2
]
S
.
G
h
o
sa
l
a
n
d
A
.
Ja
i
n
,
“
W
e
i
g
h
t
e
d
a
s
p
e
c
t
b
a
se
d
se
n
t
i
m
e
n
t
a
n
a
l
y
si
s
u
si
n
g
e
x
t
e
n
d
e
d
O
W
A
o
p
e
r
a
t
o
r
s
a
n
d
W
o
r
d
2
V
e
c
f
o
r
t
o
u
r
i
sm,
”
Mu
l
t
i
m
e
d
i
a
T
o
o
l
s
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
8
2
,
n
o
.
1
2
,
p
p
.
1
8
3
5
3
–
1
8
3
8
0
,
M
a
y
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
0
4
2
-
0
2
2
-
1
3
8
0
0
-
4.
[
3
]
D
.
A
r
i
y
u
s
,
D
.
M
a
n
o
n
g
g
a
,
a
n
d
I
.
S
e
m
b
i
r
i
n
g
,
“
E
n
h
a
n
c
i
n
g
se
n
t
i
me
n
t
a
n
a
l
y
si
s
o
f
I
n
d
o
n
e
si
a
n
t
o
u
r
i
sm
v
i
d
e
o
c
o
n
t
e
n
t
c
o
m
me
n
t
a
r
y
o
n
Ti
k
To
k
:
a
f
a
st
t
e
x
t
a
n
d
B
i
-
LS
TM
a
p
p
r
o
a
c
h
,
”
E
n
g
i
n
e
e
r
i
n
g
,
T
e
c
h
n
o
l
o
g
y
&
A
p
p
l
i
e
d
S
c
i
e
n
c
e
R
e
se
a
rc
h
,
v
o
l
.
1
4
,
n
o
.
6
,
p
p
.
1
8
0
2
0
–
1
8
0
2
8
,
D
e
c
.
2
0
2
4
,
d
o
i
:
1
0
.
4
8
0
8
4
/
e
t
a
sr
.
8
8
5
9
.
[
4
]
A
.
M
.
-
O
r
t
i
z
,
S
.
S
.
-
B
e
r
n
a
l
,
a
n
d
A
.
O
.
-
B
a
r
e
a
,
“
D
e
s
i
g
n
a
n
d
v
a
l
i
d
a
t
i
o
n
o
f
a
n
n
o
t
a
t
i
o
n
s
c
h
e
ma
s
f
o
r
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
s
i
s
i
n
t
h
e
t
o
u
r
i
s
m
s
e
c
t
o
r
,
”
I
n
f
o
rm
a
t
i
o
n
T
e
c
h
n
o
l
o
g
y
&
T
o
u
r
i
s
m
,
v
o
l
.
2
1
,
n
o
.
4
,
p
p
.
5
3
5
–
5
5
7
,
D
e
c
.
2
0
1
9
,
d
o
i
:
1
0
.
1
0
0
7
/
s
4
0
5
5
8
-
019
-
00155
-
0.
[
5
]
P
.
K
.
Ja
i
n
,
R
.
P
a
mu
l
a
,
a
n
d
G
.
S
r
i
v
a
st
a
v
a
,
“
A
s
y
s
t
e
m
a
t
i
c
l
i
t
e
r
a
t
u
r
e
r
e
v
i
e
w
o
n
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
a
p
p
l
i
c
a
t
i
o
n
s
f
o
r
c
o
n
su
m
e
r
sen
t
i
m
e
n
t
a
n
a
l
y
si
s
u
s
i
n
g
o
n
l
i
n
e
r
e
v
i
e
w
s,”
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
Re
v
i
e
w
,
v
o
l
.
4
1
,
A
u
g
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
sr
e
v
.
2
0
2
1
.
1
0
0
4
1
3
.
[
6
]
C.
-
H
.
K
u
,
Y
.
-
C
.
C
h
a
n
g
,
a
n
d
Y
.
W
a
n
g
,
“
H
o
w
t
o
s
t
r
a
t
e
g
i
c
a
l
l
y
r
e
sp
o
n
d
t
o
o
n
l
i
n
e
h
o
t
e
l
r
e
v
i
e
w
s
:
a
st
r
a
t
e
g
y
-
a
w
a
r
e
d
e
e
p
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
,
”
I
n
f
o
rm
a
t
i
o
n
& M
a
n
a
g
e
m
e
n
t
,
v
o
l
.
6
1
,
n
o
.
5
,
Ju
l
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
m
.
2
0
2
4
.
1
0
3
9
7
0
.
[
7
]
N
.
H
a
b
b
a
t
a
n
d
H
.
N
o
u
r
i
,
“
U
n
l
o
c
k
i
n
g
t
r
a
v
e
l
n
a
r
r
a
t
i
v
e
s:
a
f
u
si
o
n
o
f
st
a
c
k
i
n
g
e
n
s
e
mb
l
e
d
e
e
p
l
e
a
r
n
i
n
g
a
n
d
n
e
u
r
a
l
t
o
p
i
c
m
o
d
e
l
i
n
g
f
o
r
e
n
h
a
n
c
e
d
t
o
u
r
i
sm
c
o
mm
e
n
t
a
n
a
l
y
si
s
,
”
S
o
c
i
a
l
N
e
t
w
o
rk
An
a
l
y
si
s
a
n
d
M
i
n
i
n
g
,
v
o
l
.
1
4
,
n
o
.
1
,
A
p
r
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
3
2
7
8
-
024
-
0
1
2
5
6
-
3.
[
8
]
A
.
O
n
a
n
,
“
M
i
n
i
n
g
o
p
i
n
i
o
n
s
f
r
o
m
i
n
st
r
u
c
t
o
r
e
v
a
l
u
a
t
i
o
n
r
e
v
i
e
w
s:
a
d
e
e
p
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
,
”
C
o
m
p
u
t
e
r
A
p
p
l
i
c
a
t
i
o
n
s
i
n
E
n
g
i
n
e
e
ri
n
g
Ed
u
c
a
t
i
o
n
,
v
o
l
.
2
8
,
n
o
.
1
,
p
p
.
1
1
7
–
1
3
8
,
J
a
n
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
0
2
/
c
a
e
.
2
2
1
7
9
.
[
9
]
A
.
G
h
o
r
b
a
n
a
l
i
,
M
.
K
.
S
o
h
r
a
b
i
,
a
n
d
F
.
Y
a
g
h
mae
e
,
“
E
n
sem
b
l
e
t
r
a
n
sf
e
r
l
e
a
r
n
i
n
g
-
b
a
s
e
d
m
u
l
t
i
m
o
d
a
l
se
n
t
i
me
n
t
a
n
a
l
y
s
i
s
u
si
n
g
w
e
i
g
h
t
e
d
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s,
”
I
n
f
o
rm
a
t
i
o
n
Pr
o
c
e
ssi
n
g
&
Ma
n
a
g
e
m
e
n
t
,
v
o
l
.
5
9
,
n
o
.
3
,
M
a
y
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
p
m
.
2
0
2
2
.
1
0
2
9
2
9
.
[
1
0
]
C
.
Y
u
,
X
.
Z
h
u
,
B
.
F
e
n
g
,
L
.
C
a
i
,
a
n
d
L
.
A
n
,
“
S
e
n
t
i
m
e
n
t
a
n
a
l
y
si
s
o
f
J
a
p
a
n
e
se
t
o
u
r
i
sm
o
n
l
i
n
e
r
e
v
i
e
w
s,
”
J
o
u
r
n
a
l
o
f
D
a
t
a
a
n
d
I
n
f
o
rm
a
t
i
o
n
S
c
i
e
n
c
e
,
v
o
l
.
4
,
n
o
.
1
,
p
p
.
8
9
–
1
1
3
,
F
e
b
.
2
0
1
9
,
d
o
i
:
1
0
.
2
4
7
8
/
j
d
i
s
-
2
0
1
9
-
0
0
0
5
.
[
1
1
]
D
.
Ti
w
a
r
i
,
B
.
N
a
g
p
a
l
,
B
.
S
.
B
h
a
t
i
,
A
.
M
i
s
h
r
a
,
a
n
d
M
.
K
u
mar,
“
A
sy
s
t
e
m
a
t
i
c
r
e
v
i
e
w
o
f
s
o
c
i
a
l
n
e
t
w
o
r
k
se
n
t
i
m
e
n
t
a
n
a
l
y
si
s
w
i
t
h
c
o
m
p
a
r
a
t
i
v
e
s
t
u
d
y
o
f
e
n
se
mb
l
e
-
b
a
se
d
t
e
c
h
n
i
q
u
e
s,”
A
rt
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
R
e
v
i
e
w
,
v
o
l
.
5
6
,
n
o
.
1
1
,
p
p
.
1
3
4
0
7
–
1
3
4
6
1
,
N
o
v
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
0
4
6
2
-
023
-
1
0
4
7
2
-
w.
[
1
2
]
A
.
A
.
A
l
S
h
a
msi
a
n
d
S
.
A
b
d
a
l
l
a
h
,
“
E
n
sem
b
l
e
st
a
c
k
i
n
g
m
o
d
e
l
f
o
r
se
n
t
i
m
e
n
t
a
n
a
l
y
si
s
o
f
E
mi
r
a
t
i
a
n
d
A
r
a
b
i
c
d
i
a
l
e
c
t
s
,
”
J
o
u
rn
a
l
o
f
K
i
n
g
S
a
u
d
U
n
i
v
e
rsi
t
y
-
C
o
m
p
u
t
e
r
a
n
d
I
n
f
o
rm
a
t
i
o
n
S
c
i
e
n
c
e
s
,
v
o
l
.
3
5
,
n
o
.
8
,
S
e
p
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
k
s
u
c
i
.
2
0
2
3
.
1
0
1
6
9
1
.
[
1
3
]
A
.
M
o
h
a
mm
e
d
a
n
d
R
.
K
o
r
a
,
“
A
c
o
mp
r
e
h
e
n
si
v
e
r
e
v
i
e
w
o
n
e
n
s
e
m
b
l
e
d
e
e
p
l
e
a
r
n
i
n
g
:
o
p
p
o
r
t
u
n
i
t
i
e
s
a
n
d
c
h
a
l
l
e
n
g
e
s,
”
J
o
u
rn
a
l
o
f
K
i
n
g
S
a
u
d
U
n
i
v
e
rs
i
t
y
-
C
o
m
p
u
t
e
r
a
n
d
I
n
f
o
rm
a
t
i
o
n
S
c
i
e
n
c
e
s
,
v
o
l
.
3
5
,
n
o
.
2
,
p
p
.
7
5
7
–
7
7
4
,
F
e
b
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
k
s
u
c
i
.
2
0
2
3
.
0
1
.
0
1
4
.
[
1
4
]
M
.
P
a
o
l
a
n
t
i
e
t
a
l
.
,
“
T
o
u
r
i
s
m
d
e
s
t
i
n
a
t
i
o
n
ma
n
a
g
e
me
n
t
u
s
i
n
g
se
n
t
i
me
n
t
a
n
a
l
y
si
s
a
n
d
g
e
o
-
l
o
c
a
t
i
o
n
i
n
f
o
r
ma
t
i
o
n
:
a
d
e
e
p
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
,
”
I
n
f
o
rm
a
t
i
o
n
T
e
c
h
n
o
l
o
g
y
&
T
o
u
ri
sm
,
v
o
l
.
2
3
,
n
o
.
2
,
p
p
.
2
4
1
–
2
6
4
,
Ju
n
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
0
7
/
s
4
0
5
5
8
-
0
2
1
-
0
0
1
9
6
-
4.
[
1
5
]
B
.
A
.
P
r
a
m
o
n
o
,
“
M
u
l
t
i
l
i
n
g
u
a
l
a
sp
e
c
t
-
b
a
se
d
s
e
n
t
i
me
n
t
a
n
a
l
y
s
i
s
d
a
t
a
se
t
o
f
t
o
u
r
i
sm
r
e
v
i
e
w
s
i
n
I
n
d
o
n
e
s
i
a
(
l
a
b
e
l
e
d
w
i
t
h
G
e
m
i
n
i
2
LLM
)
,
”
Z
e
n
o
d
o
,
2
0
2
5
,
d
o
i
:
1
0
.
5
2
8
1
/
z
e
n
o
d
o
.
1
5
5
1
7
7
2
7
.
[
1
6
]
A
.
T
a
n
r
ı
se
v
d
i
,
G
.
Ö
z
t
ü
r
k
,
a
n
d
A
.
C
.
Ö
z
t
ü
r
k
,
“
A
su
p
e
r
v
i
s
e
d
d
a
t
a
m
i
n
i
n
g
a
p
p
r
o
a
c
h
f
o
r
p
r
e
d
i
c
t
i
n
g
c
o
mm
e
n
t
c
a
r
d
r
a
t
i
n
g
s,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
C
o
n
t
e
m
p
o
r
a
ry
H
o
sp
i
t
a
l
i
t
y
M
a
n
a
g
e
m
e
n
t
,
v
o
l
.
3
4
,
n
o
.
5
,
p
p
.
1
8
2
3
–
1
8
5
3
,
A
p
r
.
2
0
2
2
,
d
o
i
:
1
0
.
1
1
0
8
/
I
JC
H
M
-
05
-
2
0
2
1
-
0
6
7
5
.
[
1
7
]
S
.
H
u
a
n
g
,
L.
J
.
L
i
a
n
g
,
a
n
d
H
.
C
.
C
h
o
i
,
“
H
o
w
w
e
f
a
i
l
e
d
i
n
c
o
n
t
e
x
t
:
a
t
e
x
t
-
m
i
n
i
n
g
a
p
p
r
o
a
c
h
t
o
u
n
d
e
r
st
a
n
d
i
n
g
h
o
t
e
l
serv
i
c
e
f
a
i
l
u
r
e
s
,
”
S
u
s
t
a
i
n
a
b
i
l
i
t
y
,
v
o
l
.
1
4
,
n
o
.
5
,
F
e
b
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
su
1
4
0
5
2
6
7
5
.
[
1
8
]
S
.
R
i
y
a
n
t
o
,
I
.
S
.
S
i
t
a
n
g
g
a
n
g
,
T.
D
j
a
t
n
a
,
a
n
d
T.
D
.
A
t
i
k
a
h
,
“
C
o
mp
a
r
a
t
i
v
e
a
n
a
l
y
si
s
u
si
n
g
v
a
r
i
o
u
s p
e
r
f
o
r
ma
n
c
e
m
e
t
r
i
c
s
i
n
i
m
b
a
l
a
n
c
e
d
d
a
t
a
f
o
r
mu
l
t
i
-
c
l
a
ss
t
e
x
t
c
l
a
ssi
f
i
c
a
t
i
o
n
,
”
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
A
d
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
4
,
n
o
.
6
,
2
0
2
3
,
d
o
i
:
1
0
.
1
4
5
6
9
/
I
JA
C
S
A
.
2
0
2
3
.
0
1
4
0
6
1
1
6
.
[
1
9
]
B
.
H
a
r
j
o
,
M
u
l
j
o
n
o
,
a
n
d
R
.
A
b
d
u
l
l
a
h
,
“
A
t
t
e
n
t
i
o
n
-
b
a
se
d
s
e
n
t
e
n
c
e
e
x
t
r
a
c
t
i
o
n
f
o
r
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
w
i
t
h
i
m
p
l
i
c
i
t
a
sp
e
c
t
c
a
s
e
s
i
n
h
o
t
e
l
r
e
v
i
e
w
u
si
n
g
mac
h
i
n
e
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
m,
s
e
ma
n
t
i
c
si
m
i
l
a
r
i
t
y
,
a
n
d
B
ER
T
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
I
n
t
e
l
l
i
g
e
n
t
En
g
i
n
e
e
ri
n
g
a
n
d
S
y
st
e
m
s
,
v
o
l
.
1
6
,
n
o
.
3
,
p
p
.
1
8
9
–
2
0
0
,
J
u
n
.
2
0
2
3
,
d
o
i
:
1
0
.
2
2
2
6
6
/
i
j
i
e
s2
0
2
3
.
0
6
3
0
.
1
5
.
[
2
0
]
C.
-
C
.
W
u
,
C
.
-
W
.
Ts
a
i
,
F
.
-
E.
W
u
,
C
.
-
H
.
C
h
i
a
n
g
,
a
n
d
J.
-
C
.
C
h
i
o
u
,
“
P
l
a
n
t
a
r
p
r
e
ss
u
r
e
-
b
a
se
d
g
a
i
t
r
e
c
o
g
n
i
t
i
o
n
w
i
t
h
a
n
d
w
i
t
h
o
u
t
c
a
r
r
i
e
d
o
b
j
e
c
t
b
y
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
-
a
u
t
o
e
n
c
o
d
e
r
a
r
c
h
i
t
e
c
t
u
r
e
,
”
Bi
o
m
i
m
e
t
i
c
s
,
v
o
l
.
1
0
,
n
o
.
2
,
J
a
n
.
2
0
2
5
,
d
o
i
:
1
0
.
3
3
9
0
/
b
i
o
m
i
me
t
i
c
s1
0
0
2
0
0
7
9
.
[
2
1
]
M
.
Z
h
a
o
e
t
a
l
.
,
“
Est
a
b
l
i
s
h
me
n
t
o
f
a
D
a
q
u
g
r
a
d
e
c
l
a
ssi
f
i
c
a
t
i
o
n
m
o
d
e
l
b
a
s
e
d
o
n
c
o
mp
u
t
e
r
v
i
si
o
n
a
n
d
mac
h
i
n
e
l
e
a
r
n
i
n
g
,
”
Fo
o
d
s
,
v
o
l
.
1
4
,
n
o
.
4
,
F
e
b
.
2
0
2
5
,
d
o
i
:
1
0
.
3
3
9
0
/
f
o
o
d
s
1
4
0
4
0
6
6
8
.
[
2
2
]
L.
Y
a
o
e
t
a
l
.
,
“
R
i
b
f
r
a
c
t
u
r
e
d
e
t
e
c
t
i
o
n
s
y
st
e
m
b
a
se
d
o
n
d
e
e
p
l
e
a
r
n
i
n
g
,
”
S
c
i
e
n
t
i
f
i
c
Re
p
o
rt
s
,
v
o
l
.
1
1
,
n
o
.
1
,
D
e
c
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
3
8
/
s
4
1
5
9
8
-
021
-
0
3
0
0
2
-
7.
[
2
3
]
S
.
A
n
i
s,
S
.
S
a
a
d
,
a
n
d
M
.
A
r
e
f
,
“
S
e
n
t
i
me
n
t
a
n
a
l
y
s
i
s
o
f
h
o
t
e
l
r
e
v
i
e
w
s
u
s
i
n
g
mac
h
i
n
e
l
e
a
r
n
i
n
g
t
e
c
h
n
i
q
u
e
s,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
A
d
v
a
n
c
e
d
I
n
t
e
l
l
i
g
e
n
t
S
y
s
t
e
m
s a
n
d
I
n
f
o
rm
a
t
i
c
s
2
0
2
0
(
AI
S
I
2
0
2
0
)
,
C
h
a
m
,
S
w
i
t
z
e
r
l
a
n
d
:
S
p
r
i
n
g
e
r
,
2
0
2
1
,
p
p
.
2
2
7
–
234
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
0
3
0
-
5
8
6
6
9
-
0
_
2
1
.
[
2
4
]
M
.
B
.
A
n
l
e
y
,
G
.
S
.
N
e
g
a
sh
e
,
a
n
d
A
.
Y
.
A
y
a
l
e
w
,
“
O
p
i
n
i
o
n
m
i
n
i
n
g
o
f
t
o
u
r
i
st
s’
sen
t
i
m
e
n
t
s
:
t
o
w
a
r
d
s
a
c
o
m
p
r
e
h
e
n
si
v
e
s
e
r
v
i
c
e
i
mp
r
o
v
e
me
n
t
o
f
t
o
u
r
i
sm
i
n
d
u
st
r
y
,
”
i
n
2
0
2
3
I
EEE
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
a
t
i
o
n
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
Vi
r
t
u
a
l
En
v
i
r
o
n
m
e
n
t
s
f
o
r
Me
a
s
u
reme
n
t
S
y
st
e
m
s
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
(
C
I
V
EM
S
A)
,
Ju
n
.
2
0
2
3
,
p
p
.
1
–
6
,
d
o
i
:
1
0
.
1
1
0
9
/
C
I
V
EM
S
A
5
7
7
8
1
.
2
0
2
3
.
1
0
2
3
1
0
1
3
.
[
2
5
]
S
.
S
a
g
n
i
k
a
,
A
.
P
a
t
t
a
n
a
i
k
,
B
.
S
.
P
.
M
i
sh
r
a
,
a
n
d
S
.
K
.
M
e
h
e
r
,
“
A
r
e
v
i
e
w
o
n
mu
l
t
i
l
i
n
g
u
a
l
se
n
t
i
me
n
t
a
n
a
l
y
si
s
b
y
mac
h
i
n
e
l
e
a
r
n
i
n
g
met
h
o
d
s,
”
J
o
u
r
n
a
l
o
f
E
n
g
i
n
e
e
ri
n
g
S
c
i
e
n
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
R
e
v
i
e
w
,
v
o
l
.
1
3
,
n
o
.
2
,
p
p
.
1
5
4
–
1
6
6
,
A
p
r
.
2
0
2
0
,
d
o
i
:
1
0
.
2
5
1
0
3
/
j
e
st
r
.
1
3
2
.
1
9
.
[
2
6
]
M
.
K
.
A
n
a
m,
M
.
B
.
F
i
r
d
a
u
s,
F
.
S
u
a
n
d
i
,
L
a
t
h
i
f
a
h
,
T.
N
a
s
u
t
i
o
n
,
a
n
d
S
.
F
a
d
l
y
,
“
P
e
r
f
o
r
m
a
n
c
e
i
mp
r
o
v
e
me
n
t
o
f
mac
h
i
n
e
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
m
u
si
n
g
e
n
sem
b
l
e
m
e
t
h
o
d
o
n
t
e
x
t
m
i
n
i
n
g
,
”
i
n
2
0
2
4
I
n
t
e
rn
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
Fu
t
u
r
e
T
e
c
h
n
o
l
o
g
i
e
s
f
o
r
S
m
a
r
t
S
o
c
i
e
t
y
(
I
C
FTS
S
)
,
A
u
g
.
2
0
2
4
,
p
p
.
9
0
–
95
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
F
TSS
6
1
1
0
9
.
2
0
2
4
.
1
0
6
9
1
3
6
3
.
Evaluation Warning : The document was created with Spire.PDF for Python.