I
n
d
on
e
s
i
a
n
Jo
u
r
n
al
of
E
l
e
c
t
r
i
c
al
E
n
g
i
n
e
e
r
i
n
g a
n
d
C
om
p
u
t
e
r
S
c
i
e
n
c
e
V
ol
.
40
, N
o.
3
,
D
e
c
e
m
b
e
r
2025
, pp.
1
453
~
1
465
I
S
S
N
:
2502
-
4752, D
O
I
:
10.11591/
i
j
e
e
c
s
.
v
40
.i
3
.
pp1
453
-
1
465
1453
Jou
r
n
al
h
om
e
page
:
ht
t
p:
/
/
i
j
e
e
c
s
.i
ae
s
c
or
e
.c
om
O
p
t
i
m
i
z
i
n
g r
e
s
u
m
e
i
n
f
or
m
at
i
o
n
e
x
t
r
ac
t
i
o
n
t
h
r
o
u
g
h
T
S
H
D
se
gm
e
n
t
at
i
o
n
a
n
d
a
d
va
n
c
e
d
d
e
e
p
l
e
ar
n
i
n
g t
e
c
h
n
i
q
u
e
s
A
n
m
ar
A
b
u
h
am
d
ah
1
, M
o
h
am
m
e
d
A
l
-
S
h
ab
i
1
, S
a
n
a Jaw
ar
n
e
h
2
1
D
e
pa
r
t
m
e
n
t
of
M
a
n
a
g
e
m
e
n
t
I
n
f
o
r
m
a
t
i
o
n
S
ys
t
e
m
s
,
C
ol
l
e
g
e
of
B
us
i
n
e
s
s
A
dm
i
n
i
s
t
r
a
t
i
o
n,
T
a
i
ba
h
U
ni
ve
r
s
i
t
y
,
M
a
di
na
h
,
S
a
udi
A
r
a
bi
a
2
D
e
pa
r
t
m
e
n
t
of
C
o
m
put
e
r
S
c
i
e
n
c
e
,
A
ppl
i
e
d
C
ol
l
e
g
e
I
m
a
m
A
bd
ul
r
a
hm
a
n
B
i
n
F
a
i
s
a
l
U
n
i
ve
r
s
i
t
y
,
D
a
m
m
a
m
,
S
a
udi
A
r
a
bi
a
A
r
t
i
c
l
e
I
n
f
o
A
B
S
T
R
A
C
T
A
r
t
i
c
l
e
h
i
s
t
o
r
y
:
R
e
c
e
i
v
e
d
M
a
r
22
,
2025
R
e
v
i
s
e
d
J
ul
8
,
2025
A
c
c
e
pt
e
d
O
c
t
15
,
2025
T
h
i
s
r
e
s
e
a
r
c
h
f
o
c
u
s
e
s
o
n
a
s
i
g
n
if
i
c
a
n
t
f
a
c
to
r
i
n
th
e
n
a
tu
r
a
l
l
a
n
g
u
a
g
e
p
r
o
c
e
s
s
i
n
g
a
r
e
a
,
wh
i
c
h
i
s
e
x
tr
a
c
ti
n
g
i
n
f
o
r
m
a
ti
o
n
f
r
o
m
u
n
s
tr
u
c
tu
r
e
d
te
x
tu
a
l
d
a
ta
th
r
o
u
g
h
e
f
f
i
c
i
e
n
t
m
e
th
o
d
s
i
n
o
r
d
e
r
to
p
u
l
l
u
s
e
f
u
l
i
n
s
i
g
h
ts
a
n
d
s
tr
u
c
tu
r
e
d
r
e
p
r
e
s
e
n
tati
o
n
s
f
r
o
m
t
h
i
s
d
a
ta.
T
h
i
s
r
e
s
e
a
r
c
h
a
tt
e
m
p
ts
to
b
o
o
s
t
th
e
e
f
f
e
c
t
i
v
e
n
e
s
s
o
f
i
n
f
o
r
m
a
ti
o
n
r
e
tr
i
e
v
a
l
s
y
s
tem
s
th
r
o
u
g
h
c
o
m
p
u
tati
o
n
a
l
a
n
a
l
y
s
i
s
.
T
h
i
s
p
a
r
a
d
ig
m
i
s
e
x
p
l
o
r
e
d
i
n
th
i
s
wo
r
k
u
s
i
n
g
q
u
e
s
ti
o
n
a
n
s
we
r
i
n
g
m
o
d
e
l
s
i
n
a
n
e
x
tr
a
c
t
iv
e
s
ty
l
e
,
a
m
o
d
e
r
n
i
n
f
o
r
m
a
ti
o
n
e
x
tr
a
c
ti
o
n
a
p
p
r
o
a
c
h
,
c
r
e
a
ti
n
g
a
n
e
w
m
e
th
o
d
o
l
o
g
y
c
o
m
b
i
n
i
n
g
th
e
to
p
i
c
s
e
g
m
e
n
tati
o
n
b
a
s
e
d
o
n
h
e
a
d
i
n
g
s
d
e
tec
ti
o
n
(
T
S
HD
)
s
e
g
m
e
n
tati
o
n
a
l
g
o
r
i
th
m
a
n
d
d
e
e
p
l
e
a
r
n
i
n
g
m
e
th
o
d
s
.
T
h
e
T
S
HD
a
l
g
o
r
i
th
m
b
r
e
a
k
s
d
o
c
u
m
e
n
ts
i
n
to
s
e
c
ti
o
n
s
i
n
w
h
i
c
h
c
e
r
tai
n
to
p
i
c
s
a
r
e
a
d
d
r
e
s
s
e
d
.
R
e
f
i
n
e
d
e
x
tr
a
c
ti
o
n
m
o
d
e
l
s
a
r
e
th
e
n
u
s
e
d
to
p
r
o
c
e
s
s
th
e
s
e
d
i
s
jo
i
n
t
s
e
g
m
e
n
ts
l
e
a
d
i
n
g
to
m
o
r
e
a
c
c
u
r
a
te
a
n
d
c
o
n
tex
t
-
ju
d
i
c
i
o
u
s
e
x
tr
a
c
ti
o
n
c
o
m
p
a
r
e
d
to
n
a
i
v
e
wh
o
l
e
-
d
o
c
u
m
e
n
t
e
x
tr
a
c
ti
o
n
a
p
p
r
o
a
c
h
e
s
.
W
e
e
m
p
i
r
ica
ll
y
v
a
li
d
a
te
th
i
s
a
p
p
r
o
a
c
h
u
s
i
n
g
th
e
s
tan
f
o
r
d
q
u
e
s
ti
o
n
a
n
s
we
r
i
n
g
d
a
tas
e
t
(
S
Qu
AD
)
1
.
1
d
a
tas
e
t,
w
i
th
a
s
p
e
c
if
i
c
a
d
a
p
tati
o
n
to
r
e
s
u
m
e
s
.
E
x
p
e
r
i
m
e
n
tal
r
e
s
u
l
ts
s
h
o
w
t
h
a
t
t
h
e
p
e
r
f
o
r
m
a
n
c
e
m
e
tr
i
c
s
i
n
c
r
e
a
s
e
by
7
.
4
%
i
n
e
x
a
c
t
m
a
tch
(
E
M
)
a
n
d
b
y
7
.
8
%
i
n
F1
-
s
c
o
r
e
.
T
h
i
s
c
a
n
b
e
c
o
n
c
l
u
d
e
d
f
r
o
m
th
e
s
e
r
e
s
u
l
ts
il
l
u
s
tr
a
ti
n
g
th
e
f
e
a
s
i
b
il
i
ty
o
f
th
e
p
r
o
p
o
s
e
d
a
p
p
r
o
a
c
h
i
n
th
e
a
u
t
o
m
a
ted
i
n
f
o
r
m
a
ti
o
n
e
x
tr
a
c
ti
o
n
f
r
a
m
e
wo
r
k
s
s
u
c
h
a
s
r
e
s
u
m
e
p
r
o
c
e
s
s
i
n
g
.
K
e
y
w
or
d
s
:
D
e
e
p l
e
a
r
ni
n
g
I
nf
o
r
m
a
t
i
on
e
xt
r
a
c
t
i
on
N
a
t
ur
a
l
l
a
ng
ua
g
e
p
r
oc
e
s
s
i
ng
T
e
xt
ua
l
T
opi
c
s
e
gm
e
nt
a
t
i
on
T
r
a
ns
f
or
m
e
r
m
ode
l
s
T
h
i
s
i
s
a
n
o
p
e
n
a
c
c
e
s
s
a
r
ti
c
l
e
u
n
d
e
r
th
e
C
C
B
Y
-
SA
l
i
c
e
n
s
e
.
C
or
r
e
s
pon
di
n
g A
u
t
h
or
:
A
n
m
a
r
A
buha
m
da
h
D
e
pa
r
t
m
e
nt
of
M
a
na
ge
m
e
nt
I
nf
o
r
m
a
t
i
on
S
y
s
t
e
m
s
, C
ol
l
e
ge
of
B
us
i
ne
s
s
A
d
m
i
ni
s
t
r
a
t
i
on
T
a
i
ba
h
U
ni
v
e
r
s
i
t
y
A
l
-
M
a
di
na
h
,
Al
-
M
una
w
w
a
r
a
h,
42353, M
e
di
na
,
K
i
n
g
do
m
of
S
a
udi
A
r
a
bi
a
E
m
a
i
l
:
a
a
buha
m
da
h@
t
a
i
ba
hu.e
du.s
a
1.
I
N
T
R
O
D
U
C
T
I
O
N
M
ode
r
n
i
nf
or
m
a
t
i
on a
g
e
i
s
w
i
t
ne
s
s
i
ng
r
a
pi
d g
r
o
w
t
h of
uns
t
r
uc
t
ur
e
d t
e
xt
ua
l
da
t
a
, t
hus
m
a
ki
ng
t
e
xt
ua
l
i
nf
or
m
a
t
i
on
e
xt
r
a
c
t
i
on
a
f
unda
m
e
nt
a
l
pr
oc
e
s
s
i
n
v
a
r
i
ous
do
m
a
i
ns
i
nc
l
udi
ng
but
not
l
i
m
i
t
e
d
t
o
s
c
i
e
nt
i
f
i
c
r
e
s
e
a
r
c
h,
he
a
l
t
h
c
a
r
e
,
bus
i
ne
s
s
a
nd
e
duc
a
t
i
on.
T
hi
s
e
xpl
os
i
v
e
g
r
o
w
t
h
unde
r
s
c
or
e
s
t
he
c
r
i
t
i
c
a
l
r
ol
e
t
ha
t
na
t
ur
a
l
l
a
ng
ua
g
e
pr
o
c
e
s
s
i
ng
(
N
L
P
)
pl
a
y
s
i
n
s
ha
pi
ng
uns
t
r
uc
t
ur
e
d
da
t
a
i
nt
o
a
s
t
r
uc
t
ur
e
d
v
i
e
w
.
O
n
t
he
ot
he
r
ha
nd,
e
v
e
n i
n t
he
e
r
a
of
s
t
a
t
e
-
of
-
t
he
-
a
r
t
t
r
a
ns
f
or
m
e
r
-
ba
s
e
d
m
ode
l
s
l
i
ke
B
E
R
, c
ont
e
xt
s
pe
c
i
f
i
c
i
nf
o
r
m
a
t
i
on e
xt
r
a
c
t
i
on
i
s
di
f
f
i
c
ul
t
due
t
o
t
he
i
nhe
r
e
nt
a
m
bi
g
ui
t
i
e
s
a
nd
noi
s
e
i
n
r
e
a
l
w
or
l
d
t
e
x
t
s
[
1]
.
T
hi
s
e
xpl
os
i
on
of
uns
t
r
uc
t
ur
e
d
da
t
a
hi
g
hl
i
g
ht
s
t
he
n
e
c
e
s
s
i
t
y
f
or
s
ophi
s
t
i
c
a
t
e
d
t
e
c
hni
que
s
,
t
o
g
l
e
a
n
a
c
t
i
ona
bl
e
i
ns
i
g
ht
s
f
r
o
m
i
t
r
a
pi
dl
y.
T
ha
t
s
a
i
d,
t
he
c
o
m
pl
e
xi
t
i
e
s
of
na
t
ur
a
l
l
a
ng
ua
ge
unde
r
s
t
a
ndi
ng
a
r
e
a
s
i
g
ni
f
i
c
a
nt
ba
r
r
i
e
r
t
o
a
c
c
ur
a
t
e
pr
o
c
e
s
s
i
n
g
of
nua
nc
e
d
or
c
ont
e
xt
-
s
pe
c
i
f
i
c
i
nf
or
m
a
t
i
on.
I
n
s
pi
t
e
of
t
he
r
e
c
e
nt
s
ur
g
e
i
n
pe
r
f
or
m
a
n
c
e
a
c
r
os
s
v
a
r
i
ous
dom
a
i
ns
w
i
t
hi
n
N
L
P
br
oug
ht
f
or
t
h
by
a
d
v
a
nc
e
s
i
n
ne
w
l
y
c
onf
i
g
u
r
e
d
t
r
a
ns
f
or
m
e
r
-
b
a
s
e
d
m
ode
l
s
s
uc
h
a
s
bi
di
r
e
c
t
i
ona
l
e
nc
ode
r
r
e
pr
e
s
e
nt
a
t
i
ons
f
r
o
m
t
r
a
ns
f
or
m
e
r
s
(
B
E
R
T
)
a
nd
i
t
s
r
e
s
pe
c
t
i
v
e
d
e
r
i
v
a
t
i
v
e
s
[
1
]
,
[
2]
due
t
o
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2502
-
4752
I
ndone
s
i
a
n J
E
l
e
c
E
n
g
&
C
om
p S
c
i
, V
ol
.
40
, N
o.
3
,
D
e
c
e
m
be
r
20
25
:
1
453
-
1
46
5
1454
t
he
i
r
e
xt
r
e
m
e
s
uc
c
e
s
s
i
n f
i
e
l
ds
s
uc
h a
s
m
a
c
hi
ne
t
r
a
ns
l
a
t
i
on,
que
s
t
i
on a
ns
w
e
r
i
ng
(
QA
)
or
i
nf
or
m
a
t
i
on r
e
t
r
i
e
v
a
l
,
s
uc
h di
s
c
ov
e
r
y
m
e
t
hods
s
t
i
l
l
l
a
c
k s
pe
c
i
f
i
c
i
t
y
t
ow
a
r
ds
t
he
c
ont
e
xt
.
I
nf
o
r
m
a
t
i
on
e
xt
r
a
c
t
i
on
f
r
o
m
uns
t
r
uc
t
ur
e
d
t
e
xt
t
o
s
t
r
uc
t
ur
e
d
da
t
a
ha
s
a
l
w
a
y
s
be
e
n
a
c
or
e
pr
obl
e
m
i
n
N
L
P
,
e
v
e
r
y
now
a
nd
t
he
n
f
l
oodi
ng
a
r
a
ng
e
of
ne
w
t
e
xt
a
nd
da
t
a
i
nt
o
t
he
e
v
e
r
-
g
r
o
w
i
ng
di
g
i
t
a
l
s
pa
c
e
.
A
nu
m
b
e
r
of
m
e
t
hods
ha
v
e
be
e
n
pr
opos
e
d
t
hr
oug
h
t
he
y
e
a
r
s
,
r
a
n
g
i
n
g
f
r
o
m
t
r
a
di
t
i
ona
l
r
ul
e
-
ba
s
e
d
a
ppr
oa
c
h
e
s
t
o
i
l
l
a
g
e
-
ba
s
e
d
m
e
t
hods
ba
s
e
d
on
de
e
p
l
e
a
r
ni
ng
.
I
ni
t
i
a
l
l
y
,
e
a
r
l
y
t
e
c
hni
que
s
w
e
r
e
ba
s
e
d
on
t
he
m
a
nua
l
l
y
c
r
a
f
t
e
d
r
ul
e
s
a
nd
s
t
a
t
i
s
t
i
c
a
l
m
ode
l
s
[
2]
,
w
hi
c
h
h
a
v
e
s
ho
w
n
a
pot
e
nt
i
a
l
but
w
e
r
e
l
i
m
i
t
e
d
r
e
ga
r
di
ng
t
he
i
r
s
c
a
l
a
bi
l
i
t
y
a
nd
t
he
i
r
a
da
pt
a
bi
l
i
t
y
t
o
c
om
pl
e
x
r
e
a
l
-
w
or
l
d
doc
u
m
e
nt
s
l
i
ke
r
e
s
u
m
e
s
.
T
he
s
e
m
od
e
l
s
ha
v
e
be
e
n
e
f
f
e
c
t
i
v
e
l
y
e
xpl
oi
t
e
d
i
n
b
i
o
-
m
e
di
c
a
l
t
e
xt
m
i
ni
ng
a
nd
l
e
ga
l
doc
u
m
e
nt
a
na
l
y
s
i
s
.
B
ut
m
or
e
c
o
m
pl
e
x
a
nd
di
v
e
r
s
i
f
i
e
d
s
t
r
uc
t
ur
e
s
of
t
e
xt
a
r
e
l
e
s
s
l
i
ke
l
y
t
o
be
f
ound
i
n
t
he
m
,
e
s
pe
c
i
a
l
l
y
i
n
unr
e
s
t
r
i
c
t
e
d
doc
u
m
e
nt
s
due
t
o
dom
a
i
n i
n s
pe
c
i
f
i
c
v
oc
a
bul
a
r
y
a
nd pol
y
s
e
m
ous
w
or
ds
.
I
n
t
he
a
r
e
a
of
r
e
s
u
m
e
a
na
l
y
s
i
s
,
a
c
o
m
m
on
p
r
obl
e
m
i
s
t
o
c
or
r
e
c
t
l
y
e
xt
r
a
c
t
a
ppr
opr
i
a
t
e
i
nf
o
r
m
a
t
i
on
w
hi
l
e
e
xc
l
udi
ng
i
r
r
e
l
e
v
a
nt
i
nf
or
m
a
t
i
on.
F
or
i
ns
t
a
nc
e
,
i
n
t
he
t
a
s
k
of
e
xt
r
a
c
t
i
ng
m
e
nt
i
ons
of
e
duc
a
t
i
ona
l
i
ns
t
i
t
ut
i
on
s
f
r
om
r
e
s
u
m
e
s
,
m
ode
l
s
w
oul
d
he
r
e
a
nd
t
he
r
e
w
r
on
g
l
y
br
i
ng
out
non
-
t
a
r
g
e
t
e
nt
i
t
i
e
s
,
s
uc
h
a
s
e
m
pl
oye
r
uni
v
e
r
s
i
t
i
e
s
or
publ
i
c
a
t
i
on
v
e
nue
s
.
T
hi
s
phe
no
m
e
non
i
s
r
e
l
a
t
e
d
t
o
v
a
g
ue
ne
s
s
i
n
t
he
de
f
i
ni
t
i
on
of
c
ont
e
xt
bounda
r
i
e
s
a
nd i
t
m
a
y
a
f
f
e
c
t
pr
e
c
i
s
i
on f
or
i
nf
or
m
a
t
i
on e
xt
r
a
c
t
i
on s
y
s
t
e
m
s
.
O
ur
m
a
i
n
c
ont
r
i
but
i
on
i
s
t
o
i
m
pr
o
v
e
t
he
a
c
c
ur
a
c
y
a
nd
t
he
e
f
f
i
c
i
e
nc
y
of
r
e
s
u
m
e
c
ont
e
nt
e
xt
r
a
c
t
i
on
by
c
o
m
bi
ni
ng
t
opi
c
-
a
w
a
r
e
s
e
gm
e
nt
a
t
i
on
a
nd
de
e
p
l
e
a
r
ni
ng
-
ba
s
e
d
e
xt
r
a
c
t
i
v
e
que
s
t
i
on
a
ns
w
e
r
i
n
g
(
E
Q
A
)
.
O
ur
a
ppr
oa
c
h
l
e
v
e
r
a
ge
s
t
r
a
ns
f
or
m
e
r
-
b
a
s
e
d
Q
A
m
ode
l
s
w
i
t
h
t
he
t
opi
c
s
e
gm
e
nt
a
t
i
on
ba
s
e
d
on
he
a
di
n
g
s
de
t
e
c
t
i
on
(
T
S
H
D
)
s
e
gm
e
nt
a
t
i
on
a
l
g
or
i
t
h
m
t
o
s
e
gm
e
nt
t
he
c
og
ni
t
i
v
e
l
y
c
ohe
r
e
nt
s
e
gm
e
nt
s
a
nd
r
e
t
r
i
e
v
e
t
he
r
e
l
e
v
a
nt
i
nf
or
m
a
t
i
on
w
i
t
hout
r
e
t
r
a
i
ni
ng
t
he
m
ode
l
s
.
T
w
o
f
unda
m
e
nt
a
l
l
i
m
i
t
a
t
i
ons
c
ha
r
a
c
t
e
r
i
z
e
c
on
v
e
nt
i
ona
l
i
nf
or
m
a
t
i
on
e
xt
r
a
c
t
i
on
a
ppr
oa
c
he
s
,
w
hi
c
h
i
ndi
s
c
r
i
m
i
na
t
e
l
y
pr
oc
e
s
s
w
hol
e
doc
u
m
e
nt
s
.
T
he
f
i
r
s
t
,
t
he
y
ha
v
e
a
t
e
nde
nc
y
of
t
hr
o
w
i
n
g
t
he
pr
e
c
i
s
i
on
out
of
t
he
w
i
ndow
s
i
nc
e
t
he
y
a
r
e
bur
de
n
e
d
w
i
t
h
noi
s
e
f
r
o
m
i
r
r
e
l
e
v
a
nt
s
e
c
t
i
ons
w
hi
c
h
obs
c
ur
e
m
e
a
ni
ng
f
ul
i
ns
i
g
ht
s
.
S
e
c
ond,
t
he
s
e
m
e
t
hods
a
r
e
o
f
t
e
n
c
o
m
put
a
t
i
ona
l
l
y
i
ne
f
f
i
c
i
e
nt
,
e
s
pe
c
i
a
l
l
y
on
l
ong
doc
u
m
e
nt
s
,
be
c
a
us
e
t
he
y
w
a
s
t
e
r
e
s
our
c
e
s
a
n
a
l
y
s
i
ng
t
e
xt
t
ha
t
i
s
unl
i
ke
l
y
t
o
c
ont
r
i
but
e
t
o t
he
out
put
.
S
uc
h
i
n
de
e
d
a
r
e
t
he
w
e
a
kne
s
s
e
s
of
pr
e
v
i
ous
w
or
ks
,
w
hi
c
h
unde
r
s
c
or
e
t
he
ne
e
d f
or
m
or
e
t
a
r
ge
t
e
d a
nd
e
f
f
i
c
i
e
nt
s
ol
ut
i
ons
t
o t
he
t
a
s
k t
ha
t
c
ons
t
r
a
i
n
r
e
l
e
v
a
nt
s
e
gm
e
nt
s
of
a
doc
u
m
e
nt
a
nd
m
i
ni
m
i
z
e
r
e
dun
da
nc
y
.
I
n
or
de
r
t
o
w
o
r
k
a
r
ound
t
he
m
e
nt
i
one
d
i
s
s
ue
s
, t
hi
s
m
ode
l
pr
opos
e
s
a
ne
w
a
ppr
oa
c
h
us
i
ng
t
he
T
S
H
D
a
l
g
or
i
t
h
m
.
T
S
H
D
ha
s
t
w
o
s
t
e
ps
:
s
e
gm
e
nt
i
ng
doc
u
m
e
nt
s
i
nt
o
s
e
m
a
nt
i
c
a
l
l
y
c
ohe
r
e
nt
s
e
c
t
i
ons
,
a
nd
us
i
ng
s
e
m
a
nt
i
c
m
a
t
c
hi
n
g
t
o
i
de
nt
i
f
y
t
he
t
opi
c
a
l
r
e
l
e
v
a
n
c
e
of
t
hos
e
s
e
c
t
i
ons
t
o
s
pe
c
i
f
i
c
us
e
r
que
r
i
e
s
;
t
hi
s
a
l
l
ow
s
f
or
pr
e
c
i
s
e
a
nd
e
f
f
i
c
i
e
nt
e
xt
r
a
c
t
i
on
of
r
e
l
e
v
a
nt
i
nf
or
m
a
t
i
on.
T
hi
s
not
onl
y
he
l
ps
i
n
f
i
l
t
e
r
i
n
g
out
noi
s
e
f
r
o
m
unr
e
l
a
t
e
d
pa
r
t
s
but
a
l
s
o
e
nha
nc
e
s
t
he
us
e
of
c
o
m
put
a
t
i
ona
l
r
e
s
our
c
e
s
,
m
a
ki
ng
i
t
e
s
pe
c
i
a
l
l
y
s
ui
t
a
bl
e
f
o
r
ha
ndl
i
ng
l
a
r
ge
-
s
i
z
e
d
c
o
m
pl
e
x
pa
pe
r
s
[
3]
.
I
n
r
e
c
e
nt
r
e
s
e
a
r
c
h,
t
he
c
a
p
a
bi
l
i
t
i
e
s
of
s
e
gm
e
nt
a
t
i
on
t
e
c
hni
que
s
i
n
i
m
pr
o
v
i
n
g
i
nf
or
m
a
t
i
on
e
xt
r
a
c
t
i
on
m
ode
l
s
ha
s
be
e
n
f
ur
t
he
r
c
onf
i
r
m
e
d;
de
m
ons
t
r
a
t
i
n
g
t
ha
t
w
i
t
h
s
e
gm
e
nt
a
t
i
on,
de
e
p
l
e
a
r
ni
ng
m
ode
l
s
m
a
y
be
e
d
uc
a
t
e
d
t
o
e
xt
r
a
c
t
out
onl
y
t
he
i
nf
or
m
a
t
i
on
i
n
s
e
gm
e
nt
s
of
a
doc
u
m
e
nt
,
t
hus
r
e
a
c
hi
n
g
a
ba
l
a
nc
e
be
t
w
e
e
n hi
g
h a
c
c
ur
a
c
y
a
nd l
o
w
r
e
s
our
c
e
c
ons
u
m
pt
i
on i
n t
a
s
ks
[
4]
–
[
10]
.
I
nf
o
r
m
a
t
i
on
e
xt
r
a
c
t
i
on
s
y
s
t
e
m
s
h
a
v
e
be
c
o
m
e
m
u
c
h
m
or
e
a
c
c
ur
a
t
e
a
nd
e
f
f
i
c
i
e
nt
w
i
t
h
t
he
r
e
c
e
n
t
a
d
v
a
n
c
e
s
i
n
de
e
p
l
e
a
r
ni
ng
.
M
ode
l
s
l
i
ke
B
e
r
t
,
R
obe
r
t
a
,
a
nd
L
ong
f
or
m
e
r
ha
v
e
s
how
n
s
t
a
t
e
of
t
he
a
r
t
pe
r
f
or
m
a
nc
e
i
n pr
obl
e
m
s
l
i
ke
na
m
e
d e
nt
i
t
y
r
e
c
o
g
ni
t
i
on a
nd
E
Q
A
, e
s
pe
c
i
a
l
l
y
i
n t
he
c
a
s
e
of
s
t
r
uc
t
ur
e
d or
s
e
m
i
-
s
t
r
uc
t
ur
e
d
doc
um
e
nt
s
.
T
r
a
ns
f
or
m
e
r
s
-
ba
s
e
d
m
ode
l
s
l
i
k
e
R
oB
E
R
T
a
[
11]
,
B
E
R
T
[
2]
,
[
12]
,
L
ong
f
or
m
e
r
[
1]
,
a
n
d
X
L
N
e
t
[
13]
r
e
a
l
i
z
e
d
s
o
m
e
of
t
he
s
t
a
t
e
-
of
-
t
he
-
a
r
t
s
ol
ut
i
ons
f
or
m
a
n
y
N
L
P
t
a
s
ks
,
s
uc
h
a
s
doc
um
e
nt
c
l
a
s
s
i
f
i
c
a
t
i
on, s
e
nt
i
m
e
nt
a
n
a
l
y
s
i
s
, or
que
s
t
i
on a
ns
w
e
r
i
n
g
.
T
he
s
e
m
ode
l
s
e
xpl
oi
t
a
t
t
e
nt
i
on m
e
c
ha
ni
s
m
s
a
nd
c
a
n
m
od
e
l
l
ong
-
di
s
t
a
nc
e
de
pe
nde
nc
i
e
s
a
nd
r
e
l
a
t
i
ons
i
n
t
e
x
t
.
F
or
e
xa
m
pl
e
,
B
E
R
T
D
e
v
l
i
n
e
t
a
l
.
[
3]
f
or
t
he
f
i
r
s
t
t
i
m
e
,
a
c
hi
e
v
e
d
s
t
a
t
e
-
of
-
t
he
-
a
r
t
r
e
s
ul
t
s
on
v
a
r
i
ous
N
L
P
t
a
s
ks
,
t
hr
oug
h
l
a
r
g
e
-
s
c
a
l
e
pr
e
-
t
r
a
i
ni
ng
a
nd
f
i
ne
-
t
uni
ng
on
t
a
s
k
-
s
pe
c
i
f
i
c
da
t
a
.
R
oB
E
R
T
a
a
nd
L
on
g
f
or
m
e
r
a
r
e
pos
t
-
T
r
a
ns
f
or
m
e
r
a
r
c
hi
t
e
c
t
ur
e
s
t
ha
t
r
e
s
ol
v
e
d
ke
y
pe
r
f
or
m
a
nc
e
i
s
s
ue
s
r
a
i
s
e
d b
y
t
he
or
i
g
i
na
l
a
r
c
hi
t
e
c
t
ur
e
(
s
c
a
l
a
bi
l
i
t
y
, c
o
m
put
a
t
i
ona
l
e
f
f
i
c
i
e
nc
y)
.
O
ne
of
t
he
c
l
e
a
r
e
s
t
us
e
s
of
t
r
a
ns
f
or
m
e
r
m
ode
l
s
i
s
i
n
E
Q
A
,
w
he
r
e
t
he
obj
e
c
t
i
v
e
i
s
t
o
e
xt
r
a
c
t
r
e
l
e
v
a
nt
s
pa
ns
of
t
e
x
t
a
ns
w
e
r
i
ng
t
he
que
s
t
i
on.
E
Q
A
ha
s
t
r
a
ns
f
or
m
e
d
doc
u
m
e
nt
r
e
t
r
i
e
v
a
l
s
y
s
t
e
m
s
by
a
l
l
ow
i
ng
f
or
m
or
e
f
i
ne
-
t
une
d
e
xt
r
a
c
t
i
on
of
i
nf
or
m
a
t
i
on
f
r
o
m
s
t
r
uc
t
ur
e
d
or
s
e
m
i
-
s
t
r
uc
t
ur
e
d
t
e
xt
.
T
he
de
v
e
l
op
m
e
nt
of
s
upe
r
v
i
s
e
d
m
od
e
l
s
ha
s
he
a
v
i
l
y
r
e
l
i
e
d
on
be
nc
h
m
a
r
k
da
t
a
s
e
t
s
,
e
.
g
.,
t
he
s
t
a
nf
or
d
que
s
t
i
on
a
ns
w
e
r
i
ng
da
t
a
s
e
t
(
S
Q
uA
D
)
,
l
e
a
di
ng
t
o
g
r
e
a
t
pr
o
g
r
e
s
s
i
n
E
Q
A
.
S
Q
uA
D
1.1
[
14]
ha
s
be
c
o
m
e
t
he
s
t
a
nda
r
d
f
or
e
v
a
l
ua
t
i
on
a
nd
t
r
a
i
ni
ng
w
i
t
h
107,785
que
s
t
i
ons
,
a
nd
t
he
(
s
i
ng
l
e
)
a
ns
w
e
r
be
i
n
g
g
ua
r
a
nt
e
e
d
t
o
e
xi
s
t
i
n
t
he
pr
o
v
i
de
d
c
o
nt
e
xt
.
S
Q
uA
D
2.0
[
15]
,
w
hi
c
h
c
ont
a
i
ns
53,775
una
ns
w
e
r
a
bl
e
que
s
t
i
ons
,
ha
s
a
dde
d
a
n
a
ddi
t
i
ona
l
l
a
y
e
r
of
di
f
f
i
c
ul
t
y
a
s
m
od
e
l
s
m
us
t
no
w
e
xt
r
a
c
t
a
n
a
ns
w
e
r
or
de
t
e
r
m
i
ne
t
ha
t
a
que
s
t
i
on
doe
s
not
ha
v
e
a
v
a
l
i
d
a
ns
w
e
r
.
T
hi
s
ha
s
r
e
s
ul
t
e
d
i
n
t
he
c
r
e
a
t
i
on
of
be
t
t
e
r
m
ode
l
s
a
nd
R
oB
E
R
T
a
ha
s
e
v
e
n
be
t
t
e
r
pe
r
f
or
m
a
n
c
e
t
ha
n
B
E
R
T
due
t
o
t
he
i
m
pr
o
v
e
m
e
nt
s
i
t
br
i
ng
s
.
F
or
i
ns
t
a
nc
e
, one
m
od
e
l
a
c
hi
e
v
e
s
a
n F
1
-
s
c
or
e
o
f
85.8 on S
Q
uA
D
2.0 but
a
c
hi
e
v
e
s
a
n
F1
-
s
c
or
e
of
onl
y
66.3
on
S
Q
uA
D
1.1,
de
m
ons
t
r
a
t
i
ng
t
he
t
a
s
k
’
s
a
dde
d
c
om
pl
e
xi
t
y
.
A
l
ong
w
i
t
h S
Q
uA
D
,
ot
he
r
da
t
a
s
e
t
s
,
f
or
i
ns
t
a
nc
e
,
N
e
w
s
Q
A
,
Q
U
A
C
,
a
nd
C
ov
i
dQ
A
ha
v
e
be
e
n
r
e
m
a
r
ka
bl
y
be
ne
f
i
c
i
a
l
i
n
t
r
a
i
ni
ng
a
nd
Evaluation Warning : The document was created with Spire.PDF for Python.
I
ndone
s
i
a
n J
E
l
e
c
E
n
g
&
C
om
p S
c
i
I
S
S
N
:
2502
-
4752
O
pt
i
m
i
z
i
n
g r
e
s
um
e
i
nf
or
m
a
t
i
on e
x
t
r
ac
t
i
on t
hr
ough T
SH
D
s
e
gm
e
nt
at
i
on
an
d
…
(
A
nm
ar
A
b
u
ham
d
a
h
)
1455
e
v
a
l
ua
t
i
ng
E
Q
A
m
od
e
l
s
[
16]
–
[
18]
.
W
hi
l
e
Q
U
A
C
ha
s
14,000
di
a
l
og
ue
s
w
i
t
h
100,000
qu
e
s
t
i
ons
i
n
a
f
or
m
w
he
r
e
a
s
t
ude
nt
a
s
ke
d
W
i
ki
pe
di
a
t
e
xt
s
a
nd
t
he
t
e
a
c
he
r
a
ns
w
e
r
s
i
n
s
hor
t
t
e
xt
e
xc
e
r
pt
s
,
N
e
w
s
Q
A
i
nc
l
ude
s
119,633
que
s
t
i
on
s
ba
s
e
d
on
12,744
ne
w
s
a
r
t
i
c
l
e
s
.
C
o
v
i
dQ
A
,
b
y
c
ont
r
a
s
t
,
c
ons
i
s
t
s
of
2,019
que
s
t
i
on
-
a
ns
w
e
r
pa
i
r
s
dr
a
w
n
f
r
o
m
147
a
r
t
i
c
l
e
s
a
bout
C
O
V
I
D
-
19.
F
1
-
s
c
or
e
i
s
s
t
i
l
l
one
of
t
he
m
a
i
n
m
e
t
r
i
c
s
f
or
e
v
a
l
ua
t
i
n
g
m
od
e
l
s
t
r
a
i
ne
d on s
uc
h
da
t
a
s
e
t
s
l
i
ke
i
n
T
a
bl
e
1
[
19]
.
T
a
bl
e
1 s
ho
w
s
a
c
o
m
p
a
r
i
s
on of
t
he
F1
-
s
c
or
e
of
a
v
a
r
i
e
t
y
of
t
r
a
ns
f
or
m
e
r
s
on
s
e
v
e
r
a
l
da
t
a
s
e
t
s
w
hi
c
h
s
how
s
p
e
r
f
or
m
a
nc
e
di
f
f
e
r
e
nc
e
s
on
t
he
ba
s
i
s
of
t
r
a
i
ni
ng
s
i
z
e
,
que
s
t
i
on c
l
a
r
i
t
y
a
nd l
e
v
e
l
of
c
ont
e
xt
c
o
m
pl
e
xi
t
y
.
T
a
bl
e
1. C
o
m
pa
r
a
t
i
v
e
p
e
r
f
or
m
a
nc
e
of
t
r
a
ns
f
or
m
e
r
-
b
a
s
e
d
m
ode
l
s
a
c
r
os
s
da
t
a
s
e
t
s
(
F1
-
s
c
or
e
)
M
ode
l
S
Q
uA
D
2.
0
N
e
w
s
Q
A
Q
U
A
C
C
ovi
dQ
A
X
L
N
e
t
B
A
S
E
64.
9
53.
2
30.
1
44.
9
B
E
R
T
B
A
S
E
64.
7
52.
1
28.
6
44.
8
R
O
B
E
R
T
A
B
A
S
E
68.
2
57.
0
31.
3
44.
5
A
L
B
E
R
T
B
A
S
E
64.
8
51.
8
19.
5
42.
4
C
on
v
B
e
r
t
B
A
S
E
67.
4
55.
7
31.
5
44.
9
B
A
R
T
B
A
S
E
67.
6
56.
2
29.
1
45.
3
B
E
R
T
B
A
S
E
-
B
I
L
S
T
M
65.
0
52.
6
28.
9
45.
6
I
n
f
a
c
t
,
m
os
t
of
t
he
e
f
f
o
r
t
s
i
n
t
hi
s
a
r
e
a
a
r
e
ba
s
e
d
on
m
od
e
l
s
t
ha
t
ope
r
a
t
e
a
t
t
he
doc
u
m
e
n
t
l
e
v
e
l
t
o
e
xt
r
a
c
t
i
nf
or
m
a
t
i
on. H
o
w
e
v
e
r
, pa
pe
r
s
, dr
ug
l
e
a
f
l
e
t
s
, W
i
ki
pe
di
a
a
r
t
i
c
l
e
s
, a
nd r
e
s
u
m
é
s
t
y
pi
c
a
l
l
y
i
nc
l
ude
di
s
t
i
nc
t
pa
r
t
s
t
ha
t
a
ddr
e
s
s
s
pe
c
i
f
i
c
m
a
t
t
e
r
s
[
20]
–
[2
2
]
.
D
i
f
f
e
r
e
nt
e
v
a
l
ua
t
i
on
r
e
s
ul
t
s
c
a
n
be
c
a
us
e
d
b
y
(
i
)
be
i
n
g
a
f
f
e
c
t
e
d
by
t
r
a
i
ni
ng
s
e
t
s
i
z
e
,
(
ii
)
s
i
m
pl
i
c
i
t
y
o
f
t
he
que
s
t
i
on
-
a
nd
-
a
ns
w
e
r
f
or
m
ul
a
t
i
ons
a
nd
(
iii
)
t
e
xt
ua
l
c
ont
e
xt
c
o
m
pl
e
xi
t
y
[
20]
.
F
or
i
ns
t
a
nc
e
,
da
t
a
s
e
t
s
w
hi
c
h
a
r
e
w
e
l
l
-
s
t
r
uc
t
ur
e
d,
s
uc
h
a
s
S
Q
uA
D
2.0,
a
t
t
a
i
n
a
m
uc
h
l
a
r
ge
r
F1
-
s
c
or
e
(
85.8)
c
om
p
a
r
e
d
t
o
S
Q
uA
D
1.1
(
66.3)
,
i
ndi
c
a
t
i
ng
t
ha
t
t
he
s
t
r
uc
t
ur
e
of
t
he
doc
um
e
nt
ha
s
a
n
i
m
pa
c
t
on t
he
pe
r
f
or
m
a
n
c
e
of
t
he
m
od
e
l
.
T
o a
l
l
e
v
i
a
t
e
t
he
di
f
f
i
c
ul
t
y i
n pr
oc
e
s
s
i
ng
l
on
g
doc
u
m
e
nt
s
,
g
r
e
a
t
e
r
m
ode
l
s
l
i
ke
l
ong
f
or
m
e
r
h
a
v
e
s
ho
w
n
pr
o
m
i
s
i
ng
r
e
s
ul
t
s
i
n
t
e
xt
-
ba
s
e
d
a
ns
w
e
r
e
xt
r
a
c
t
i
on
t
a
s
ks
by
us
i
ng
a
s
c
a
l
a
bl
e
a
t
t
e
nt
i
on
f
or
m
ul
a
t
i
on
[
11]
.
6R
oB
E
R
T
a
R
oB
E
R
T
a
i
s
a
pi
one
e
r
i
n
t
hi
s
c
a
t
e
g
or
y
–
i
t
t
a
ke
s
t
he
c
l
a
s
s
i
c
B
E
R
T
a
r
c
hi
t
e
c
t
ur
e
a
nd
pe
r
f
or
m
s
s
om
e
m
a
s
s
i
v
e
upgr
a
de
s
on
i
t
.
B
ut
t
he
s
e
m
ode
l
s
a
r
e
r
e
s
our
c
e
-
i
nt
e
ns
i
v
e
t
o
be
a
bl
e
t
o
us
e
on a
r
e
s
our
c
e
-
l
i
m
i
t
e
d
de
v
i
c
e
.
F
or
c
o
m
put
a
t
i
ona
l
l
y
e
f
f
i
c
i
e
nt
a
l
t
e
r
n
a
t
i
v
e
s
,
D
i
s
t
i
l
B
E
R
T
i
s
a
di
s
t
i
l
l
e
d
v
a
r
i
a
nt
of
B
E
R
T
w
hi
c
h
pr
o
v
i
d
e
l
a
r
ge
s
pe
e
dups
a
nd r
e
duc
e
d s
i
z
e
w
i
t
h l
i
t
t
l
e
qua
l
i
t
y
l
os
s
[
20]
.
H
o
w
e
v
e
r
,
t
he
s
e
m
ode
l
s
a
r
e
s
t
i
l
l
not
s
ui
t
a
bl
e
f
or
uns
t
r
uc
t
ur
e
d
or
o
v
e
r
l
ooke
d
da
t
a
,
i
ndi
c
a
t
i
ng
t
he
r
e
qui
r
e
m
e
nt
of
no
v
e
l
pr
e
-
pr
oc
e
s
s
i
n
g
m
e
t
hods
. T
opi
c
s
e
gm
e
nt
a
t
i
on a
nd i
de
nt
i
f
i
c
a
t
i
on c
a
n he
l
p t
he
i
nf
or
m
a
t
i
on
e
xt
r
a
c
t
i
on
a
nd
s
um
m
a
r
y
t
a
s
ks
,
be
c
a
us
e
t
he
y
a
l
l
ow
m
od
e
l
s
t
o
f
oc
us
on
t
he
r
e
l
e
v
a
nt
pe
r
i
p
he
r
y
.
H
i
g
h
qua
l
i
t
y
s
e
gm
e
nt
a
t
i
ons
a
r
e
t
he
r
e
f
or
e
ne
e
de
d
t
o
pr
oc
e
s
s
na
t
ur
a
l
l
a
ng
ua
g
e
t
a
s
ks
e
f
f
e
c
t
i
v
e
l
y
i
n
t
he
s
e
r
e
g
i
m
e
s
.
T
he
s
e
t
r
a
de
-
of
f
s
s
how
t
ha
t
m
ode
l
s
ha
v
e
t
o
be
s
e
l
e
c
t
e
d
a
c
c
or
di
n
g
t
o
t
he
a
ppl
i
c
a
t
i
on
-
s
pe
c
i
f
i
c
r
e
qui
r
e
m
e
nt
s
[
3
]
,
[
23]
.
C
r
os
s
-
doc
um
e
nt
s
e
gm
e
nt
a
t
i
on
i
s
a
t
e
c
hni
que
f
or
s
pl
i
t
t
i
ng
up
a
doc
u
m
e
nt
i
nt
o
s
m
a
l
l
e
r
,
m
o
r
e
m
a
na
g
e
a
bl
e
pi
e
c
e
s
, t
o a
ddr
e
s
s
r
e
a
di
ng
c
o
m
pl
e
xi
t
y
i
s
s
ue
s
f
a
c
e
d
w
i
t
h l
a
r
g
e
r
, c
o
m
pl
e
x doc
u
m
e
nt
s
.
O
ne
s
uc
h
r
e
c
e
nt
a
l
g
or
i
t
h
m
be
l
ong
s
t
o
t
hi
s
g
e
ne
r
a
l
f
r
a
m
e
w
o
r
k:
t
he
T
S
H
D
t
ha
t
di
v
i
de
s
doc
u
m
e
nt
s
i
nt
o c
ohe
r
e
nt
s
e
c
t
i
ons
by
t
opi
c
. F
or
r
e
t
r
i
e
v
a
l
t
a
s
k T
S
H
D
c
a
n a
c
hi
e
v
e
hi
g
he
r
p
e
r
f
or
m
a
nc
e
s
i
nc
e
i
t
r
e
t
a
i
ns
t
hos
e
s
e
gm
e
nt
s
w
hos
e
m
e
a
ni
ng
i
s
c
ohe
r
e
nt
a
nd
c
ons
i
s
t
e
nt
.
T
hi
s
a
l
l
ow
s
a
m
or
e
f
oc
us
s
e
d
a
nd
s
t
r
uc
t
ur
e
d
e
xt
r
a
c
t
i
on
m
e
t
hod
t
o
pr
oc
e
e
d
ot
he
r
t
ha
n
j
us
t
pr
oc
e
s
s
i
ng
w
hol
e
doc
u
m
e
nt
s
.
T
hi
s
t
i
t
l
i
ng
of
w
or
k
i
s
i
l
l
us
t
r
a
t
e
d
by
T
S
H
D
f
or
r
e
s
u
m
e
pr
oc
e
s
s
i
ng
a
nd
l
e
g
a
l
doc
u
m
e
nt
pr
oc
e
s
s
i
ng
.
F
or
i
ns
t
a
nc
e
,
T
S
H
D
a
c
hi
e
v
e
d
a
n
F
1
-
s
c
or
e
of
a
r
ound
96%
a
nd
a
v
e
r
y
l
o
w
s
e
gm
e
nt
a
t
i
on
e
r
r
o
r
r
a
t
e
of
a
ppr
oxi
m
a
t
e
l
y
2%
[
2
3
]
t
hus
,
i
n
t
hi
s
s
e
ns
e
i
t
c
a
n
be
s
e
e
n
a
s
a
us
e
f
ul
t
ool
t
o
i
m
pr
o
v
e
i
nf
or
m
a
t
i
on
e
xt
r
a
c
t
i
on
m
ode
l
s
.
T
hi
s
v
a
r
i
a
bi
l
i
t
y
i
s
r
e
duc
e
d
w
he
n
us
i
ng
T
S
H
D
a
l
g
or
i
t
h
m
on
r
e
s
u
m
e
t
e
xt
s
a
s
t
he
a
l
g
or
i
t
h
m
s
e
gm
e
nt
s
doc
um
e
nt
s
t
o
t
op
i
c
c
ohe
r
e
nt
s
ubs
e
t
s
r
e
duc
i
ng
t
he
e
f
f
e
c
t
of
t
he
uns
t
r
uc
t
ur
e
d
na
t
ur
e
of
t
he
t
e
xt
a
nd
t
he
pr
e
s
e
nc
e
of
dom
a
i
n
de
pe
nde
nt
j
a
r
g
on,
t
he
r
e
by
l
e
t
t
i
ng
m
ode
l
s
l
i
ke
L
ong
f
o
r
m
e
r
t
o f
oc
us
on i
m
p
r
o
v
e
d
w
i
ndow
of
c
ont
e
xt
, r
a
t
he
r
t
ha
n t
he
uns
t
r
uc
t
ur
e
d
r
a
w
t
e
xt
.
S
e
v
e
r
a
l
l
i
t
e
r
a
t
ur
e
w
or
ks
ha
s
e
m
pha
s
i
z
e
d
on
a
ut
o
m
a
t
e
d
e
xt
r
a
c
t
i
on
of
us
e
f
ul
da
t
a
f
r
o
m
r
e
s
um
e
s
us
i
n
g
N
L
P
a
ppl
i
c
a
t
i
ons
i
n t
he
f
i
e
l
d of
r
e
s
um
e
a
n
a
l
y
s
i
s
. M
e
t
hods
s
uc
h a
s
ke
yw
or
d e
xt
r
a
c
t
i
on, e
nt
i
t
y
r
e
c
o
g
ni
t
i
on, a
nd
s
y
nt
a
c
t
i
c
pa
r
s
i
ng
ha
v
e
be
e
n
a
ppl
i
e
d
f
o
r
i
de
nt
i
f
y
i
ng
i
nf
or
m
a
t
i
on
s
uc
h
a
s
s
k
i
l
l
s
,
qua
l
i
f
i
c
a
t
i
o
ns
a
nd
e
xpe
r
i
e
nc
e
.
H
o
w
e
v
e
r
,
s
uc
h
s
i
m
pl
e
s
t
r
a
t
e
g
i
e
s
c
a
n
f
a
c
e
i
s
s
ue
s
i
n
a
r
e
a
l
s
c
e
na
r
i
o
c
ont
e
xt
w
he
r
e
t
he
r
e
s
u
m
e
s
a
r
e
hi
gh
di
m
e
ns
i
ona
l
a
nd uns
t
r
uc
t
ur
e
d or
t
he
r
e
m
a
y be
a
d
v
a
n
c
e
d bi
o do
m
a
i
n kno
w
l
e
d
g
e
i
n t
he
r
e
s
u
m
e
.
T
hi
s
ha
s
l
e
d t
o
t
he
r
e
c
e
nt
e
f
f
or
t
of
i
nt
e
gr
a
t
i
ng
s
e
gm
e
nt
a
t
i
on
a
l
g
or
i
t
h
m
s
s
uc
h
a
s
T
S
H
D
w
i
t
h
pow
e
r
f
ul
de
e
p
l
e
a
r
ni
ng
m
od
e
l
s
t
ha
t
not
onl
y
he
l
p
t
o i
m
pr
o
v
e
t
he
a
c
c
ur
a
c
y
of
t
he
r
e
s
u
m
e
i
nf
or
m
a
t
i
on
e
xt
r
a
c
t
i
on,
but
a
l
s
o
d
e
m
ons
t
r
a
t
e
a
hi
g
h
g
e
ne
r
a
l
i
z
a
t
i
on a
b
i
l
i
t
y
f
or
t
he
out
-
of
-
do
m
a
i
n e
xa
m
pl
e
s
.
T
hi
s
r
e
s
e
a
r
c
h
pr
e
s
e
nt
s
a
ne
w
f
r
a
m
e
w
o
r
k
c
o
m
bi
ni
ng
t
he
T
S
H
D
s
e
gm
e
nt
a
t
i
on
a
l
g
or
i
t
h
m
w
i
t
h
D
L
-
ba
s
e
d
e
xt
r
a
c
t
i
v
e
Q
A
m
ode
l
s
t
o
i
m
pr
o
v
e
t
he
qua
l
i
t
y
of
i
nf
or
m
a
t
i
on
e
xt
r
a
c
t
i
on
f
r
o
m
r
e
s
u
m
e
s
.
S
uc
h
a
hy
br
i
d
m
e
t
hod
be
ne
f
i
t
s
f
r
om
t
he
a
d
v
a
nt
a
ge
s
of
bot
h
s
e
gm
e
nt
a
t
i
on
a
nd
que
s
t
i
on
-
a
ns
w
e
r
i
ng
a
p
pr
oa
c
he
s
s
i
nc
e
i
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2502
-
4752
I
ndone
s
i
a
n J
E
l
e
c
E
n
g
&
C
om
p S
c
i
, V
ol
.
40
, N
o.
3
,
D
e
c
e
m
be
r
20
25
:
1
453
-
1
46
5
1456
e
ns
ur
e
s
a
m
or
e
t
a
r
g
e
t
e
d
a
nd
c
ont
e
xt
-
r
e
l
e
v
a
nt
m
a
nne
r
of
da
t
a
e
xt
r
a
c
t
i
on.
B
a
s
e
d
on
t
he
r
e
s
ul
t
s
of
our
e
xpe
r
i
m
e
nt
a
l
s
t
udy
,
w
hi
c
h
i
n
v
ol
v
e
d
t
he
us
e
of
t
he
S
Q
uA
D
1.1
i
n
t
he
c
ont
e
xt
of
a
r
e
s
um
e
,
i
t
i
s
pos
s
i
bl
e
t
o
c
onc
l
ude
t
he
f
e
a
s
i
bi
l
i
t
y
of
t
he
pr
opos
e
d
f
r
a
m
e
w
o
r
k,
a
s
e
v
i
de
nc
e
d
b
y
t
he
i
nc
r
e
a
s
e
i
n
e
xa
c
t
m
a
t
c
h
(
E
M
)
s
c
or
e
s
a
nd
F1
-
s
c
or
e
s.
T
hi
s
r
e
s
e
a
r
c
h
i
s
s
pe
c
i
f
i
c
a
l
l
y
f
oc
us
e
d
on
r
e
s
um
e
s
,
w
hi
c
h
a
r
e
a
m
ong
t
he
bui
l
di
ng
bl
oc
ks
of
hi
r
i
ng
.
R
e
s
um
e
s
c
ont
a
i
n
di
f
f
e
r
e
nt
s
e
c
t
i
ons
(
e
.g
.,
e
duc
a
t
i
on,
pr
of
e
s
s
i
ona
l
e
xpe
r
i
e
nc
e
,
s
ki
l
l
s
)
,
w
hi
c
h
r
e
qui
r
e
pr
e
c
i
s
e
e
xt
r
a
c
t
i
on
m
e
c
ha
ni
s
m
s
i
n
or
de
r
t
o
e
ns
ur
e
r
e
l
i
a
bl
e
c
a
ndi
da
t
e
e
v
a
l
u
a
t
i
on.
T
hus
,
t
he
m
a
i
n
pu
r
pos
e
of
t
hi
s
w
or
k
i
s
t
o
e
s
t
a
bl
i
s
h
a
uni
f
i
e
d
f
r
a
m
e
w
or
k
m
e
r
g
i
ng
t
r
a
ns
f
or
m
e
r
m
ode
l
s
w
i
t
h
T
S
H
D
ba
s
e
d
t
opi
c
s
e
gm
e
nt
a
t
i
on.
T
he
s
e
gm
e
nt
a
t
i
on
i
n t
hi
s
f
r
a
m
e
w
or
k
r
e
du
c
e
s
i
nf
or
m
a
t
i
on
e
xt
r
a
c
t
i
on
t
o
pr
oc
e
s
s
i
ng
c
ont
e
xt
u
a
l
l
y
s
m
a
l
l
r
e
l
e
v
a
nt
s
e
gm
e
nt
s
onl
y
w
i
t
hout
t
he
r
e
qui
r
e
m
e
nt
of
r
e
-
t
r
a
i
ni
ng
m
ode
l
s
f
or
do
c
u
m
e
nt
l
e
v
e
l
d
a
t
a
.
T
hi
s
da
t
a
i
s
t
he
n
f
or
m
a
t
t
e
d
i
nt
o
t
a
bl
e
s
a
nd
i
s
us
e
d
i
n dow
ns
t
r
e
a
m
c
o
m
put
a
t
i
ons
m
a
ki
ng
t
he
de
c
i
s
i
on
pi
pe
l
i
n
e
s
f
a
s
t
e
r
a
nd
m
or
e
e
f
f
i
c
i
e
nt
.
N
o
v
e
l
t
e
xt
s
u
m
m
a
r
i
z
a
t
i
on
a
nd
e
xt
r
a
c
t
i
on
f
r
a
m
e
w
o
r
ks
ha
v
e
e
m
e
r
g
e
d
r
e
c
e
nt
l
y
e
xp
l
oi
t
i
ng
pr
e
-
t
r
a
i
ne
d
l
a
ng
ua
g
e
m
ode
l
s
,
w
hi
c
h
a
l
l
ow
t
he
m
t
o
pr
oduc
e
a
c
c
ur
a
t
e
a
nd
e
f
f
i
c
i
e
nt
out
put
,
e
s
pe
c
i
a
l
l
y
f
or
he
t
e
r
o
g
e
ne
ous
doc
u
m
e
nt
s
e
c
t
i
ons
, e
.g
.
,
r
e
s
u
m
e
s
[
9]
, [
10]
.
I
n
t
hi
s
pa
pe
r
,
w
e
pr
opos
e
t
o
t
a
c
kl
e
t
he
s
e
c
ha
l
l
e
nge
s
w
i
t
h
a
ne
w
f
r
a
m
e
w
or
k
t
ha
t
c
o
m
bi
n
e
s
t
opi
c
-
a
w
a
r
e
s
e
gm
e
nt
a
t
i
on
w
i
t
h
t
r
a
ns
f
or
m
e
r
-
b
a
s
e
d
m
od
e
l
s
.
T
hi
s
w
or
k
e
xt
e
nds
pr
e
v
i
ous
w
o
r
k
i
n
t
he
a
r
e
a
of
doc
u
m
e
nt
s
e
gm
e
nt
a
t
i
on
a
nd
de
e
p
l
e
a
r
ni
n
g
E
Q
A
.
P
r
o
m
i
ne
nt
w
o
r
ks
of
T
S
H
D
f
or
t
opi
c
-
a
w
a
r
e
s
e
gm
e
nt
a
t
i
on
[
3]
,
R
oB
E
R
T
a
f
or
r
obus
t
l
a
ng
u
a
ge
m
od
e
l
l
i
ng
[
11]
,
a
nd
L
on
g
f
or
m
e
r
f
or
l
on
g
-
doc
u
m
e
nt
pr
oc
e
s
s
i
ng
[
3]
h
a
v
e
pr
o
v
i
d
e
d i
ns
pi
r
a
t
i
on f
or
our
w
o
r
k.
H
o
w
e
v
e
r
,
m
a
n
y
pr
e
v
i
ous
a
ppr
o
a
c
he
s
t
e
nd t
o ha
ndl
e
t
he
e
nt
i
r
e
doc
u
m
e
nt
i
n
a
v
e
r
y
s
t
r
a
i
g
ht
f
or
w
a
r
d
m
a
nne
r
,
w
hi
c
h
i
n
t
ur
n
hur
t
s
t
he
pr
e
di
c
t
i
on
a
c
c
ur
a
c
y
a
nd
r
unt
i
m
e
e
s
t
i
m
a
t
i
on.
W
e
bui
l
d
on
t
he
s
e
w
i
t
h
our
m
ode
l
,
w
hi
c
h
a
dds
t
opi
c
-
a
w
a
r
e
s
e
gm
e
nt
a
t
i
on
t
o
t
r
a
ns
f
or
m
e
r
-
ba
s
e
d
Q
A
m
od
e
l
s
,
e
na
bl
i
ng
m
or
e
a
c
c
ur
a
t
e
c
on
t
e
x
t
s
e
l
e
c
t
i
o
n a
n
d f
oc
us
e
d e
x
t
r
a
c
t
i
o
n.
O
n
t
he
o
t
he
r
ha
nd
,
t
he
m
a
i
n
nov
e
l
t
i
e
s
of
t
hi
s
w
or
k a
r
e
:
-
A
no
v
e
l
c
o
m
bi
na
t
i
on of
T
S
H
D
s
e
gm
e
nt
a
t
i
on a
l
g
or
i
t
h
m
a
nd e
xt
r
a
c
t
i
v
e
Q
A
m
ode
l
s
f
or
r
e
s
um
e
pr
e
s
e
nt
a
t
i
o
n
i
m
pr
o
v
e
m
e
nt
.
-
E
m
pi
r
i
c
a
l
e
v
a
l
ua
t
i
on
of
t
he
pr
opos
e
d
m
e
t
hod
w
i
t
h
a
da
pt
e
d
S
Q
uA
D
1.1
que
s
t
i
ons
s
how
i
ng
s
i
g
ni
f
i
c
a
nt
pe
r
f
or
m
a
nc
e
i
m
p
r
o
v
e
m
e
nt
s
i
n bot
h
EM
a
nd
F
1.
-
E
v
i
de
nc
e
t
ha
t
t
he
c
om
put
a
t
i
ona
l
c
os
t
i
s
l
e
s
s
e
ne
d
by
c
ont
e
xt
f
i
l
t
e
r
i
ng
,
w
i
t
hout
ne
c
e
s
s
i
t
y
t
o
r
e
t
r
a
i
n
m
ode
l
s
on s
e
gm
e
nt
e
d da
t
a
.
T
he
s
e
c
ont
r
i
but
i
ons
a
r
e
c
ons
i
s
t
e
nt
l
y
s
how
n
t
hr
oug
hout
t
he
m
e
t
hod,
e
m
pi
r
i
c
a
l
e
v
a
l
ua
t
i
on,
a
nd
di
s
c
us
s
i
on
s
e
c
t
i
ons
i
n t
h
i
s
pa
pe
r
.
T
hi
s
w
or
k
i
s
m
ot
i
v
a
t
e
d
by
t
he
pr
obl
e
m
of
e
xt
r
a
c
t
i
ng
s
t
r
uc
t
ur
e
d
i
nf
or
m
a
t
i
on
f
r
o
m
uns
t
r
uc
t
ur
e
d
r
e
s
u
m
e
s
due
t
o t
he
g
r
o
w
i
ng
r
e
qui
r
e
m
e
nt
of
e
f
f
i
c
i
e
nt
a
ut
o
m
a
t
e
d r
e
c
r
ui
t
m
e
nt
s
y
s
t
e
m
s
.
A
l
t
houg
h N
L
P
ha
s
m
a
d
e
c
ons
i
de
r
a
bl
e
pr
o
g
r
e
s
s
,
c
ur
r
e
nt
m
e
t
hods
do
not
us
ua
l
l
y
de
a
l
w
e
l
l
w
i
t
h
t
he
h
e
t
e
r
o
g
e
ne
ou
s
noi
s
y
na
t
ur
e
o
f
r
e
s
u
m
e
d
a
t
a
.
I
n
t
hi
s
pa
pe
r
,
w
e
a
t
t
e
m
pt
t
o
a
ddr
e
s
s
t
he
a
bo
v
e
i
s
s
ue
by
i
nt
r
oduc
i
ng
a
no
v
e
l
hy
br
i
d
f
r
a
m
e
w
o
r
k
w
hi
c
h
i
nt
e
g
r
a
t
e
s
t
he
T
S
H
D
s
e
gm
e
nt
a
t
i
on
m
ode
l
w
i
t
h
t
r
a
ns
f
or
m
e
r
-
ba
s
e
d
Q
A
m
ode
l
.
I
n
t
hi
s
pa
pe
r
,
w
e
s
t
ud
y
w
he
t
he
r
w
e
c
a
n
e
nha
nc
e
t
he
a
c
c
ur
a
c
y
of
e
xt
r
a
c
t
i
on
by
t
opi
c
-
a
w
a
r
e
s
e
gm
e
nt
a
t
i
on
a
t
t
he
l
ow
c
o
m
put
a
t
i
ona
l
c
os
t
.
T
he
w
or
k
i
s
c
l
os
e
t
o
t
he
f
oc
us
of
t
he
j
our
na
l
,
na
m
e
l
y
a
d
v
a
n
c
e
d
c
o
m
put
i
ng
t
e
c
hni
que
s
a
nd
t
he
i
r
a
ppl
i
c
a
t
i
ons
i
n r
e
a
l
-
l
i
f
e
dom
a
i
ns
, i
n t
hi
s
c
a
s
e
i
n t
he
a
r
e
a
of
i
nf
or
m
a
t
i
on e
xt
r
a
c
t
i
on a
nd doc
u
m
e
nt
pr
oc
e
s
s
i
ng
.
T
he
pa
pe
r
i
s
or
g
a
ni
z
e
d
a
s
f
ol
l
ow
s
:
S
e
c
t
i
on
2
pr
e
s
e
nt
s
t
he
pr
opo
s
e
d
a
ppr
oa
c
h
t
he
hy
br
i
d
pi
pe
l
i
ne
of
t
r
a
ns
f
or
m
e
r
ba
s
e
d
E
Q
A
a
nd
T
S
H
D
.
S
e
c
t
i
on
3
c
ont
a
i
ns
t
he
r
e
s
ul
t
s
a
nd
di
s
c
us
s
i
on
,
i
n
w
hi
c
h
w
e
e
v
a
l
ua
t
e
t
he
pe
r
f
or
m
a
nc
e
of
t
he
p
r
opos
e
d
f
r
a
m
e
w
o
r
k
a
g
a
i
ns
t
s
o
m
e
ba
s
e
l
i
ne
m
ode
l
s
.
C
onc
l
udi
ng
r
e
m
a
r
ks
a
r
e
g
i
v
e
n
i
n
s
e
c
t
i
on
4
i
n
w
hi
c
h
w
e
pr
e
s
e
nt
a
l
i
s
t
of
m
e
r
i
t
s
of
t
he
r
e
s
e
a
r
c
h
t
o
t
he
de
v
e
l
op
m
e
nt
of
e
f
f
i
c
i
e
nt
a
nd
c
ont
e
xt
-
or
i
e
nt
e
d i
nf
or
m
a
t
i
on e
xt
r
a
c
t
i
on s
y
s
t
e
m
. R
e
f
e
r
e
nc
e
s
, c
ons
t
r
a
i
nt
s
, i
m
pl
i
c
a
t
i
ons
a
nd f
ut
ur
e
s
t
u
dy
di
r
e
c
t
i
on.
2.
M
E
T
H
O
D
I
n t
hi
s
s
e
c
t
i
on w
e
d
e
s
c
r
i
be
t
he
da
t
a
s
e
t
us
e
d i
n t
hi
s
s
t
udy
, t
he
e
v
a
l
ua
t
i
on
m
e
t
r
i
c
s
us
e
d, ho
w
t
he
T
S
H
D
a
l
g
or
i
t
h
m
w
or
ks
,
a
nd
t
he
i
nf
or
m
a
t
i
on
e
xt
r
a
c
t
i
on
m
ode
l
s
t
ha
t
ha
v
e
be
e
n
us
e
d
t
o
m
e
a
s
ur
e
t
he
e
f
f
e
c
t
i
v
e
ne
s
s
of
our
pr
opos
e
d
m
e
t
hod.
M
or
e
o
v
e
r
,
a
f
ur
t
he
r
de
s
c
r
i
pt
i
on
of
m
e
t
hodol
og
i
c
a
l
c
ons
i
de
r
a
t
i
ons
t
h
a
t
unde
r
pi
nne
d
t
he
e
v
a
l
ua
t
i
on pr
oc
e
s
s
t
o c
onduc
t
a
t
hor
oug
h a
nd r
obus
t
a
na
l
y
s
i
s
i
s
doc
um
e
nt
e
d.
2
.
1
. D
at
as
e
t
d
e
s
c
r
i
p
t
i
on
W
e
c
ol
l
e
c
t
105
r
e
s
um
e
s
i
n
E
ng
l
i
s
h
f
r
om
onl
i
ne
por
t
f
ol
i
os
,
s
e
r
v
i
ng
a
s
our
da
t
a
s
e
t
.
T
he
s
e
r
e
s
um
e
s
c
o
m
e
i
n
a
l
l
t
y
pe
s
of
f
or
m
a
t
s
a
nd
ha
v
e
a
v
a
r
i
e
t
y
of
s
t
r
uc
t
ur
e
s
t
o
s
ui
t
y
our
ne
e
ds
.
W
e
m
a
nua
l
l
y
l
a
be
l
e
d
i
m
por
t
a
nt
f
i
e
l
ds
(
i
.e
.
,
na
m
e
,
t
e
l
e
phone
nu
m
be
r
,
e
m
a
i
l
,
e
du
c
a
t
i
on)
t
o
m
a
i
nt
a
i
n
c
ons
i
s
t
e
nc
y
.
S
i
nc
e
r
e
s
u
m
e
s
a
r
e
uns
t
r
uc
t
ur
e
d t
e
xt
f
i
l
e
s
, t
he
y
a
l
l
ha
v
e
di
f
f
e
r
e
nt
t
e
m
pl
a
t
e
s
a
nd f
or
m
a
t
s
i
n s
how
i
ng
i
nf
or
m
a
t
i
on
w
i
t
h a
s
i
g
ni
f
i
c
a
nt
di
f
f
e
r
e
nc
e
i
n
t
he
num
be
r
of
s
e
c
t
i
ons
a
nd
t
he
i
r
or
de
r
.
T
he
r
e
i
s
a
l
s
o
no
c
ons
i
s
t
e
nc
y
of
f
ont
t
y
pe
,
s
i
z
e
or
c
ol
our
,
w
hi
c
h
a
dd
t
o
t
he
s
t
r
uc
t
ur
a
l
non
-
ho
m
o
g
e
ne
i
t
y
of
t
he
da
t
a
s
e
t
.
T
h
e
da
t
a
s
e
t
f
r
a
m
e
w
or
k
u
s
e
d
t
o
a
s
s
e
s
s
t
he
pr
opos
e
d
i
nf
or
m
a
t
i
on
e
xt
r
a
c
t
i
on
m
e
t
hod
w
a
s
ba
s
e
d
on
t
he
S
Q
uA
D
1.1,
w
hi
c
h
w
a
s
i
n
t
r
oduc
e
d
i
n
[
14]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
ndone
s
i
a
n J
E
l
e
c
E
n
g
&
C
om
p S
c
i
I
S
S
N
:
2502
-
4752
O
pt
i
m
i
z
i
n
g r
e
s
um
e
i
nf
or
m
a
t
i
on e
x
t
r
ac
t
i
on t
hr
ough T
SH
D
s
e
gm
e
nt
at
i
on
an
d
…
(
A
nm
ar
A
b
u
ham
d
a
h
)
1457
T
he
a
s
s
e
s
s
m
e
nt
c
ons
i
s
t
e
d
of
a
s
ki
ng
c
e
r
t
a
i
n
que
s
t
i
ons
t
o
g
e
t
t
he
i
m
por
t
a
nt
de
t
a
i
l
s
out
of
t
he
r
e
s
u
m
e
s
,
i
n
t
he
w
a
y
g
i
v
e
n be
l
o
w
:
-
E
xt
r
a
c
t
i
ng
t
he
r
e
s
u
m
e
o
w
ne
r
’
s
na
m
e
,
a
nd t
he
que
s
t
i
on:
w
ha
t
i
s
m
y
f
ul
l
na
m
e
?
-
E
xt
r
a
c
t
t
he
e
m
a
i
l
a
ddr
e
s
s
,
a
nd t
he
que
s
t
i
on:
w
ha
t
i
s
t
he
e
m
a
i
l
a
ddr
e
s
s
?
-
O
bt
a
i
ni
ng
t
he
m
obi
l
e
phone
num
be
r
,
a
nd t
he
que
s
t
i
on:
w
ha
t
i
s
t
he
m
obi
l
e
phone
nu
m
be
r
?
-
E
xt
r
a
c
t
i
ng
t
he
ba
c
he
l
or
’
s
m
a
j
or
,
a
nd t
he
que
s
t
i
on:
f
or
w
hi
c
h
m
a
j
or
i
s
t
he
ba
c
he
l
or
’
s
de
gr
e
e
?
-
A
n
e
xt
r
a
c
t
or
,
w
hi
c
h
i
s
us
e
d
t
o
e
xt
r
a
c
t
t
he
uni
v
e
r
s
i
t
y
a
w
a
r
di
ng
t
he
ba
c
h
e
l
or
,
a
nd
t
he
que
s
t
i
o
n
:
f
r
om
w
hi
c
h
uni
v
e
r
s
i
t
y
ba
c
he
l
or
de
g
r
e
e
?
I
t
s
houl
d
be
not
e
d,
t
he
444
que
s
t
i
ons
i
n
t
he
t
e
s
t
da
t
a
s
e
t
a
l
l
ha
v
e
c
or
r
e
s
pondi
ng
a
ns
w
e
r
s
.
N
ow
,
t
he
S
Q
uA
D
da
t
a
s
e
t
ha
s
t
he
f
ol
l
ow
i
ng
s
t
r
uc
t
ur
e
d f
o
r
m
a
t
:
-
I
d:
U
ni
que
i
de
nt
i
f
i
e
r
a
s
s
i
g
ne
d f
or
e
a
c
h of
t
he
q
ue
s
t
i
on
.
-
T
i
t
l
e
:
t
he
na
m
e
of
t
he
doc
u
m
e
nt
f
i
l
e
f
or
t
he
r
e
s
u
m
e
.
-
C
ont
e
xt
:
t
he
t
e
xt
s
e
gm
e
nt
t
ha
t
t
he
a
ns
w
e
r
s
houl
d be
e
xt
r
a
c
t
e
d f
r
o
m
.
-
Q
ue
s
t
i
on:
w
ha
t
y
ou ne
e
d t
o r
e
t
r
i
e
v
e
t
hi
s
i
nf
or
m
a
t
i
on
.
-
A
ns
w
e
r
s
:
c
or
r
e
c
t
a
ns
w
e
r
(
s
)
,
w
i
t
h up t
o 3 pos
s
i
bl
e
v
a
l
i
d a
ns
w
e
r
s
f
or
e
a
c
h
que
s
t
i
on
.
T
he
de
f
i
ne
d
s
y
s
t
e
m
a
t
i
c
a
ppr
oa
c
h
he
l
ps
t
o
pr
ov
i
de
a
c
ons
i
s
t
e
nt
e
v
a
l
ua
t
i
on
m
e
t
hodol
og
y
f
or
de
t
e
r
m
i
ni
ng
t
he
s
ui
t
a
bi
l
i
t
y
of
i
nf
or
m
a
t
i
on e
xt
r
a
c
t
i
on
m
ode
l
s
t
o di
f
f
e
r
e
nt
a
nd pot
e
nt
i
a
l
l
y
uns
t
r
uc
t
ur
e
d r
e
s
u
m
e
l
a
y
out
s
.
2
.
2
. E
val
u
at
i
o
n
m
e
t
r
i
c
s
W
e
e
v
a
l
ua
t
e
i
nf
or
m
a
t
i
on
e
xt
r
a
c
t
i
on
m
ode
l
s
by
t
w
o
w
i
de
l
y
us
e
d
m
e
t
r
i
c
s
EM
a
nd
F1
-
s
c
or
e
.
T
he
s
e
e
v
a
l
ua
t
i
on m
e
t
r
i
c
s
qua
nt
i
f
y
ho
w
a
c
c
ur
a
t
e
l
y
t
he
m
ode
l
r
e
t
r
i
e
v
e
s
t
he
a
ns
w
e
r
f
r
o
m
t
e
xt
u
a
l
da
t
a
[
2]
, [
14]
.
a)
E
v
a
l
ua
t
e
s
w
he
t
he
r
t
he
a
ns
w
e
r
pr
e
di
c
t
e
d
by
t
he
m
ode
l
v
e
r
b
a
t
i
m
m
a
t
c
he
s
t
he
g
r
ound
t
r
ut
h
a
ns
w
e
r
.
T
he
m
e
t
r
i
c
g
i
v
e
s
a
y
e
s
/
no
s
c
or
e
:
A
s
c
or
e
of
100
i
s
a
w
a
r
d
e
d
i
f
t
he
pr
e
di
c
t
e
d
a
ns
w
e
r
i
s
i
de
nt
i
c
a
l
t
o
t
he
a
c
t
ua
l
a
ns
w
e
r
.
O
t
he
r
w
i
s
e
,
i
t
i
s
s
c
or
e
d 0.
T
he
o
v
e
r
a
l
l
E
M
s
c
or
e
i
s
c
a
l
c
ul
a
t
e
d by
a
v
e
r
a
g
i
n
g
t
he
E
M
s
c
o
r
e
s
o
v
e
r
e
a
c
h e
v
a
l
ua
t
i
on que
s
t
i
on [
2
4
].
b)
F1
-
s
c
or
e
m
e
t
r
i
c
c
o
m
put
e
s
t
he
i
nt
e
r
s
e
c
t
i
on of
w
or
ds
i
n t
he
c
a
ndi
da
t
e
a
ns
w
e
r
s
ni
ppe
t
a
nd t
he
g
r
ound t
r
ut
h
s
ni
ppe
t
. I
t
i
s
c
om
put
e
d us
i
ng
pr
e
c
i
s
i
on a
nd r
e
c
a
l
l
, w
hi
c
h
c
a
n
be
de
f
i
ne
d a
s
:
P
r
e
c
i
s
i
on:
t
he
num
be
r
of
c
or
r
e
c
t
l
y
pr
e
di
c
t
e
d
w
o
r
ds
di
v
i
de
d
b
y
t
he
num
be
r
of
w
or
ds
i
n
t
he
pr
e
di
c
t
e
d
a
ns
w
e
r
.
R
e
c
a
l
l
:
n
um
be
r
of
w
or
ds
t
ha
t
w
e
r
e
pr
e
di
c
t
e
d
a
c
c
ur
a
t
e
l
y
i
n
c
o
m
pa
r
i
s
on
w
i
t
h
num
b
e
r
of
w
o
r
ds
i
n
a
c
t
ua
l
a
ns
w
e
r
.
U
s
i
ng
t
he
(
1
)
,
w
e
c
o
m
put
e
t
he
F1
-
s
c
or
e
.
1
=
2
×
×
+
(
1)
T
he
n,
t
he
o
v
e
r
a
l
l
F1
-
s
c
or
e
a
c
r
os
s
a
l
l
e
v
a
l
ua
t
i
on
que
s
t
i
ons
i
s
a
n
a
v
e
r
a
g
e
of
t
he
F1
-
s
c
or
e
s
t
o
g
i
v
e
a
g
e
ne
r
a
l
i
de
a
o
f
t
he
m
od
e
l
pe
r
f
or
m
a
n
c
e
.
H
o
w
e
v
e
r
,
t
hi
s
m
e
t
r
i
c
a
l
l
ow
s
f
o
r
a
m
or
e
p
r
e
c
i
s
e
e
va
l
ua
t
i
on
t
ha
t
t
a
ke
s
i
nt
o
a
c
c
ount
pa
r
t
i
a
l
c
or
r
e
c
t
ne
s
s
of
t
he
pr
e
di
c
t
e
d
a
ns
w
e
r
w
he
n
i
t
c
ont
a
i
ns
s
om
e
,
but
not
a
l
l
w
or
ds
pr
e
s
e
nt
i
n
t
he
a
c
t
ua
l
a
ns
w
e
r
.
2
.
3
. G
e
n
e
r
al
m
e
c
h
an
i
s
m
of
t
h
e
T
S
H
D
al
gor
i
t
h
m
an
d
i
t
s
op
e
r
at
i
on
s
t
age
s
T
S
H
D
a
l
g
or
i
t
h
m
c
ons
i
s
t
s
of
t
hr
e
e
m
a
i
n
s
t
a
ge
s
t
o
e
xt
r
a
c
t
i
nf
or
m
a
t
i
on
f
r
o
m
uns
t
r
uc
t
ur
e
d
t
e
xt
da
t
a
.
T
he
s
e
s
t
e
ps
a
r
e
t
a
ke
n
t
o
f
a
c
i
l
i
t
a
t
e
t
he
s
y
s
t
e
m
a
t
i
c
pe
r
f
or
m
a
nc
e
of
s
e
gm
e
nt
a
t
i
on
ba
s
e
d
on
pr
e
-
pr
oc
e
s
s
i
ng
,
he
a
di
ng
de
t
e
c
t
i
on, c
ont
e
nt
e
xt
r
a
c
t
i
on, a
nd s
o on:
a)
T
he
i
n
i
t
i
a
l
s
t
e
p
of
t
he
a
l
g
or
i
t
hm
i
s
on
pr
e
-
pr
oc
e
s
s
i
ng
,
w
hi
c
h
r
e
f
e
r
s
t
o
t
he
t
r
a
ns
f
or
m
a
t
i
on
a
n
d
r
e
f
i
ne
m
e
nt
of
r
a
w
t
e
xt
ua
l
da
t
a
f
or
s
ub
s
e
q
ue
n
t
pr
oc
e
s
s
i
ng
. T
hi
s
pha
s
e
i
s
c
om
pr
i
s
e
d
of
a
s
e
q
ue
nc
e
of
s
uc
c
e
s
s
i
v
e
s
t
e
p
s
:
T
e
xt
pr
e
-
pr
oc
e
s
s
i
ng
:
b
a
s
i
c
t
r
a
ns
f
or
m
a
t
i
on of
t
e
xt
.
L
i
ne
s
t
oke
ni
z
a
t
i
o
n:
d
i
v
i
di
ng
t
e
xt
i
nt
o m
ul
t
i
pl
e
l
i
ne
s
f
or
pr
e
c
i
s
e
a
na
l
y
s
i
s
.
D
a
t
a
pr
e
-
pr
oc
e
s
s
i
ng
:
c
l
e
a
ni
n
g
unde
s
i
r
a
bl
e
c
ha
r
a
c
t
e
r
s
, s
pa
c
e
s
, e
xt
r
a
s
ym
bol
s
.
N
or
m
a
l
i
z
a
t
i
on
:
a
d
j
u
s
t
i
ng
t
e
x
t
f
or
m
a
t
s
t
o
a
u
ni
f
or
m
s
t
a
nda
r
d,
s
uc
h a
s
c
a
s
e
nor
m
a
l
i
z
a
t
i
o
n
a
nd f
ont
s
t
y
l
e
s
.
E
num
e
r
a
t
i
on
of
l
i
ne
s
:
i
de
nt
i
f
y
i
ng
e
a
c
h l
i
ne
of
t
e
xt
w
i
t
h uni
que
l
i
ne
num
be
r
s
.
L
i
ne
s
r
e
f
i
ne
m
e
nt
:
r
e
f
o
r
m
a
t
t
i
ng
of
t
he
t
e
xt
i
n or
de
r
t
o
m
a
i
nt
a
i
n c
ohe
r
e
nc
e
a
nd
i
m
p
r
o
v
e
s
e
gm
e
nt
a
t
i
on
T
he
s
e
s
t
e
ps
e
ns
ur
e
t
he
i
nc
r
e
a
s
e
d
qua
l
i
t
y
of
t
he
d
a
t
a
o
v
e
r
a
l
l
,
l
e
a
di
ng
t
o
l
e
s
s
e
r
di
s
c
r
e
pa
nc
i
e
s
a
s
t
he
da
t
a
pr
oc
e
e
ds
t
o t
he
f
ol
l
ow
i
ng
pha
s
e
.
b)
T
he
he
a
di
ng
s
de
t
e
c
t
i
on
s
t
a
g
e
,
a
t
t
h
i
s
s
t
a
g
e
t
he
a
l
g
or
i
t
hm
f
i
nd
s
s
e
c
t
i
on
he
a
d
i
ng
s
w
i
t
hi
n
t
he
d
oc
um
e
nt
a
nd
br
i
ng
s
c
on
s
i
s
t
e
nc
y
t
o
t
he
v
a
r
i
a
t
i
o
ns
of
he
a
d
i
ng
t
e
r
m
s
. T
hi
s
c
o
ns
i
s
t
s
of
t
w
o l
i
ne
a
r
s
c
a
n
s
one
a
f
t
e
r
a
no
t
he
r
:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2502
-
4752
I
ndone
s
i
a
n J
E
l
e
c
E
n
g
&
C
om
p S
c
i
, V
ol
.
40
, N
o.
3
,
D
e
c
e
m
be
r
20
25
:
1
453
-
1
46
5
1458
C
ue
w
or
ds
s
c
a
n:
l
ooks
f
or
c
om
m
onl
y
us
e
d
ke
yw
or
ds
t
ha
t
a
r
e
c
o
m
m
onl
y
f
ound
i
n
s
e
c
t
i
on
he
a
de
r
s
(
l
i
ke
“
E
duc
a
t
i
on,”
“
E
xpe
r
i
e
nc
e
”
)
C
ue
phr
a
s
e
s
s
c
a
n:
i
de
nt
i
f
y
i
ng
m
ul
t
i
-
w
or
d
e
xpr
e
s
s
i
ons
a
nd
he
a
de
r
v
a
r
i
a
t
i
ons
t
o
c
o
m
bi
ne
s
e
c
t
i
on
na
m
e
s
a
c
r
os
s
di
v
e
r
s
e
doc
u
m
e
nt
t
y
pe
s
.
c)
B
y
l
a
be
l
l
i
ng
s
e
c
t
i
on
he
a
di
ng
s
,
t
he
a
l
g
or
i
t
h
m
e
ns
ur
e
s
g
r
e
a
t
e
r
c
ons
i
s
t
e
nc
y
i
n
s
e
gm
e
nt
i
de
nt
i
f
i
c
a
t
i
on,
w
hi
c
h i
m
pr
o
v
e
s
t
he
s
ubs
e
que
nt
s
e
gm
e
nt
a
t
i
on pe
r
f
or
m
a
nc
e
.
d)
T
he
l
a
s
t
s
t
a
g
e
i
s
s
e
gm
e
nt
a
t
i
on
w
hi
c
h
e
xt
r
a
c
t
s
c
ont
e
nt
unde
r
e
a
c
h
he
a
de
r
a
nd
f
or
m
a
t
s
i
t
i
n
a
m
a
c
hi
ne
-
r
e
a
da
bl
e
f
or
m
a
t
.
S
e
c
t
i
ons
a
r
e
be
i
ng
e
xt
r
a
c
t
e
d
a
s
ke
y
-
v
a
l
ue
pa
i
r
s
,
w
he
r
e
e
a
c
h
s
e
c
t
i
on
he
a
di
n
g
w
i
l
l
be
t
he
ke
y
,
a
nd
t
he
c
or
r
e
s
pondi
ng
c
ont
e
nt
w
i
l
l
be
t
he
v
a
l
ue
i
n
J
S
O
N
f
or
m
a
t
.
B
y
s
t
r
uc
t
ur
i
n
g
t
he
i
nf
or
m
a
t
i
on
i
n
s
uc
h
a
w
a
y
,
i
t
be
c
o
m
e
s
f
a
r
m
or
e
e
f
f
i
c
i
e
nt
f
or
do
w
ns
t
r
e
a
m
p
r
oc
e
s
s
e
s
t
o
w
or
k
w
i
t
h
i
t
,
a
l
l
ow
i
ng
f
or
be
t
t
e
r
que
r
y
i
n
g
t
o i
nf
or
m
a
t
i
on a
nd a
ut
o
m
a
t
i
ng
t
he
a
n
a
l
y
s
i
s
a
s
w
e
l
l
.
I
ns
t
e
a
d,
r
e
w
or
ki
n
g
doc
u
m
e
nt
s
t
hr
oug
h
t
he
s
e
t
hr
e
e
s
t
a
g
e
s
e
na
bl
e
s
t
he
T
S
H
D
a
l
g
or
i
t
h
m
t
o
i
de
nt
i
f
y
t
e
xt
ua
l
i
nf
or
m
a
t
i
on
m
or
e
a
c
c
ur
a
t
e
l
y
a
nd e
f
f
e
c
t
i
v
e
, e
s
p
e
c
i
a
l
l
y
i
n c
i
r
c
u
m
s
t
a
nc
e
s
of
uns
t
r
uc
t
ur
e
d da
t
a
.
2
.
4
. U
s
e
d
I
n
f
or
m
at
i
on
e
xt
r
ac
t
i
on
m
od
e
l
s
T
he
m
e
c
ha
ni
s
m
w
a
s
i
m
pl
e
m
e
nt
e
d
a
nd
e
v
a
l
ua
t
e
d
on
t
hr
e
e
f
i
ne
-
t
une
d
i
nf
or
m
a
t
i
on
e
xt
r
a
c
t
i
o
n
m
ode
l
s
t
r
a
i
ne
d on r
e
s
u
m
e
da
t
a
us
i
ng
t
he
P
yT
o
r
c
h
l
i
br
a
r
y
.
T
he
m
ode
l
s
us
e
d i
n t
hi
s
s
t
udy
a
r
e
de
s
c
r
i
b
e
d a
s
f
ol
l
ow
s
.
2
.
4.1
.
A
u
t
ot
r
ai
n
-
r
e
s
u
m
e
_p
ar
s
e
r
m
od
e
l
F
or
m
ode
l
e
v
a
l
ua
t
i
o
n,
w
e
c
ho
s
e
t
hr
e
e
f
i
ne
-
t
une
d
E
Q
A
m
ode
l
s
t
r
a
i
ne
d on
r
e
s
um
e
s
.
T
he
f
i
r
s
t
m
ode
l
w
e
us
e
i
s
t
he
A
ut
ot
r
a
i
n
-
r
e
s
um
e
_pa
r
s
e
r
w
hi
c
h
i
s
b
ui
l
t
on
t
he
L
ong
f
or
m
e
r
m
ode
l
t
ha
t
c
a
n
e
f
f
i
c
i
e
nt
l
y
ha
nd
l
e
l
ong
doc
um
e
nt
. T
h
i
s
m
ode
l
i
s
ba
s
e
d
on
t
he
l
ong
f
or
m
e
r
a
r
c
hi
t
e
c
t
ur
e
[
1]
,
w
hi
c
h
s
c
a
l
e
s
s
e
l
f
-
a
t
t
e
nt
i
on
--
a
s
pe
c
i
a
l
c
a
s
e
of
t
he
m
ul
t
i
-
he
a
d
a
t
t
e
n
t
i
on
m
e
c
ha
ni
s
m
us
e
d i
n
t
he
t
r
a
ns
f
or
m
e
r
a
r
c
hi
t
e
c
t
ur
e
--
l
i
ne
a
r
l
y
w
i
t
h
r
e
s
p
e
c
t
t
o t
he
l
e
ng
t
h
of
t
he
i
np
ut
s
e
que
nc
e
,
a
nd
a
l
l
ow
s
pr
oc
e
s
s
i
ng
l
o
ng
doc
um
e
nt
s
m
or
e
e
f
f
i
c
i
e
n
t
l
y
t
ha
n
r
e
g
ul
a
r
t
r
a
ns
f
or
m
e
r
-
ba
s
e
d
m
ode
l
s
.
T
he
ba
s
e
bl
oc
k
s
of
t
he
m
ode
l
a
r
e
v
i
s
ua
l
i
z
e
d
i
n
T
a
bl
e
2,
m
odi
f
i
e
d
i
n
a
w
a
y
t
o
ob
t
a
i
n
m
a
xi
m
u
m
pe
r
f
o
r
m
a
nc
e
g
i
v
e
n
t
o
t
he
t
a
s
k
of
e
xt
r
a
c
t
i
ng
s
t
r
uc
t
ur
e
d
i
nf
or
m
a
t
i
on
f
r
om
r
e
s
um
e
s
.
I
n
a
ddi
t
i
on
,
a
ut
ot
r
a
i
n
-
r
e
s
um
e
_pa
r
s
e
r
m
ode
l
i
n m
or
e
l
ow
-
l
e
v
e
l
s
i
de
us
e
s
a
s
e
q
ue
nc
e
l
e
ng
t
h
of
4
09
6 t
oke
n
s
w
i
t
h a
t
t
e
n
t
i
on w
i
n
dow
i
ng
t
o
pr
oc
e
s
s
l
ong
r
e
s
um
e
s
qu
i
c
k
l
y
w
i
t
h c
on
t
e
x
t
i
nt
a
c
t
,
a
s
s
h
ow
n i
n
T
a
bl
e
2.
T
a
bl
e
2. C
onf
i
g
ur
a
t
i
on de
t
a
i
l
s
of
t
he
a
ut
ot
r
a
i
n
-
r
e
s
um
e
_pa
r
s
e
r
m
od
e
l
ba
s
e
d on l
ong
f
or
m
e
r
a
r
c
hi
t
e
c
t
ur
e
F
e
a
t
ur
e
N
a
m
e
V
a
l
ue
D
e
s
c
r
i
pt
i
o
n
H
i
dde
n_a
c
t
g
e
l
u
H
i
dde
n
l
a
ye
r
a
c
t
i
va
t
i
o
n
f
u
n
c
t
i
o
n
H
i
dde
n_s
i
z
e
768
H
i
dde
n
l
a
ye
r
s
i
z
e
(
e
m
be
ddi
ng
di
m
e
n
s
i
o
n
s
)
M
a
x
_pos
i
t
i
o
n_e
m
bi
ddi
ng
s
4098
M
a
x
i
m
u
m
pos
i
t
i
o
n
e
m
be
ddi
ng
s
l
i
m
i
t
M
ode
l
_t
y
pe
l
ongf
or
m
e
r
M
ode
l
t
ype
N
um
_a
t
t
e
n
t
i
o
n
_
h
e
a
ds
12
N
um
be
r
of
a
t
t
e
n
t
i
o
n
h
e
a
ds
N
um
_
h
i
dde
n
_l
a
ye
r
s
12
N
um
be
r
of
h
i
dde
n
l
a
ye
r
s
V
oc
a
b_s
i
z
e
50265
V
oc
a
bul
a
r
y
s
i
z
e
M
a
x
_l
e
ng
t
h
384
M
a
x
i
m
u
m
i
nput
l
e
ng
t
h
f
ot
t
h
e
m
ode
l
2
.4.
2
.
C
V
_C
u
s
t
om
_D
S
m
od
e
l
T
hi
s
i
s
ba
s
e
d
on
[
4]
,
R
oB
E
R
T
a
:
a
r
obus
t
l
y
opt
i
m
i
z
e
d
B
E
R
T
p
r
e
-
t
r
a
i
ni
ng
a
ppr
oa
c
h
.
U
nl
i
ke
B
E
R
T
,
R
oB
E
R
T
a
f
i
xe
s
s
o
m
e
of
i
t
s
f
r
a
c
t
i
ona
l
i
m
pr
o
v
e
m
e
nt
:
a)
R
e
m
o
v
a
l
of
t
he
ne
xt
s
e
nt
e
nc
e
pr
e
di
c
t
i
on
(
N
S
P
)
t
a
s
k
t
o
l
e
t
t
he
m
ode
l
t
r
a
i
n
onl
y
on
t
oke
n
-
l
e
v
e
l
a
nd
s
e
nt
e
nc
e
-
l
e
v
e
l
r
e
pr
e
s
e
nt
a
t
i
ons
w
i
t
hout
i
nt
e
r
-
s
e
nt
e
nc
e
de
pe
nde
nc
i
e
s
a
f
f
e
c
t
i
ng
r
e
pr
e
s
e
nt
a
t
i
on
l
e
a
r
ni
ng
.
b)
D
y
na
m
i
c
m
a
s
ki
ng
dur
i
n
g
t
r
a
i
ni
ng
,
w
hi
c
h i
nt
r
oduc
e
s
v
a
r
i
a
bi
l
i
t
y
of
m
a
s
ke
d t
oke
ns
dur
i
n
g
pr
e
-
t
r
a
i
ni
ng
f
o
r
m
o
r
e
e
f
f
e
c
t
i
v
e
m
ode
l
t
r
a
i
ni
ng
.
c)
F
ur
t
he
r
pr
e
-
t
r
a
i
ni
ng
on
l
a
r
g
e
r
a
nd
l
on
g
e
r
t
e
xt
s
e
que
nc
e
s
t
o
a
l
l
ow
t
hi
s
m
od
e
l
t
o
l
e
a
r
n
r
i
c
h
e
r
c
ont
e
xt
ua
l
r
e
pr
e
s
e
nt
a
t
i
ons
a
nd g
e
n
e
r
a
l
i
z
e
w
e
l
l
a
c
r
os
s
a
dow
ns
t
r
e
a
m
of
N
L
P
t
a
s
ks
.
T
he
a
dopt
e
d
m
od
e
l
ha
s
s
om
e
di
f
f
e
r
e
nc
e
s
w
i
t
h
B
E
R
T
on
s
e
v
e
r
a
l
s
e
t
t
i
ng
s
(
T
a
bl
e
3)
t
o
t
r
a
ns
f
or
m
i
t
i
nt
o
a
m
ode
l
f
or
i
nf
or
m
a
t
i
on
e
xt
r
a
c
t
i
on
t
a
s
ks
[
2
5
]
.
I
n
c
ont
r
a
s
t
t
o
B
E
R
T
,
t
hi
s
R
o
B
E
R
T
a
m
o
de
l
e
l
i
m
i
na
t
e
s
t
h
e
N
S
P
t
a
s
k
s
o
t
ha
t
i
t
c
a
n
c
onc
e
nt
r
a
t
e
on
t
he
t
oke
n
-
l
e
v
e
l
a
nd
s
e
nt
e
nc
e
-
l
e
v
e
l
r
e
pr
e
s
e
nt
a
t
i
ons
w
i
t
hout
c
ons
i
de
r
a
t
i
on
of
i
nt
e
r
f
e
r
e
nc
e
f
r
o
m
i
nt
e
r
-
s
e
nt
e
nc
e
l
e
a
r
ni
ng
.
I
t
a
l
s
o
us
e
s
d
y
na
m
i
c
m
a
s
ki
ng
dur
i
ng
t
r
a
i
ni
ng
a
n
d
pr
e
-
t
r
a
i
ne
d
on
l
a
r
g
e
r
t
e
xt
s
e
que
nc
e
s
w
hi
c
h
i
m
pr
o
v
e
s
i
t
s
c
ont
e
xt
unde
r
s
t
a
ndi
ng
a
nd
ge
ne
r
a
l
i
z
a
t
i
on
a
c
r
os
s
v
a
r
i
e
d
r
e
s
u
m
e
l
a
y
out
s
.
T
a
bl
e
3:
C
V
_C
us
t
om
_D
S
m
ode
l
our
C
V
_C
u
s
t
om
_D
S
m
ode
l
bu
i
l
ds
on
B
E
R
T
by
e
l
i
m
i
na
t
i
ng
t
he
N
S
P
t
a
s
k
a
nd
u
s
i
ng
c
us
t
om
i
z
a
bl
e
dy
na
m
i
c
m
a
s
ki
n
g
dur
i
ng
t
r
a
i
ni
ng
t
o
i
m
pr
o
v
e
ge
ne
r
a
l
i
z
a
t
i
on
on di
f
f
e
r
e
nt
r
e
s
u
m
e
f
or
m
a
t
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
ndone
s
i
a
n J
E
l
e
c
E
n
g
&
C
om
p S
c
i
I
S
S
N
:
2502
-
4752
O
pt
i
m
i
z
i
n
g r
e
s
um
e
i
nf
or
m
a
t
i
on e
x
t
r
ac
t
i
on t
hr
ough T
SH
D
s
e
gm
e
nt
at
i
on
an
d
…
(
A
nm
ar
A
b
u
ham
d
a
h
)
1459
T
a
bl
e
3.
A
r
c
hi
t
e
c
t
ur
a
l
a
nd t
r
a
i
ni
ng
c
onf
i
g
ur
a
t
i
ons
of
t
he
C
V
_C
us
t
om
_D
S
m
ode
l
ba
s
e
d on
R
oB
E
R
T
a
F
e
a
t
ur
e
n
a
m
e
V
a
l
ue
D
e
s
c
r
i
pt
i
o
n
H
i
dde
n
-
act
G
e
l
u
H
i
dde
n
l
a
ye
r
a
c
t
i
va
t
i
o
n
f
u
n
c
t
i
o
n
H
i
dde
n_s
i
z
e
768
H
i
dde
n
l
a
ye
r
s
i
z
e
(
e
m
be
ddi
ng
di
m
e
n
s
i
o
n
s
)
M
a
x
_
pos
i
t
i
o
n_e
m
be
ddi
ng
s
514
M
a
x
i
m
u
m
pos
i
t
i
o
n
e
m
be
ddi
ng
s
l
i
m
i
t
M
ode
l
_t
y
pe
R
obe
r
t
a
M
ode
l
t
ype
N
um
_a
t
t
e
n
t
i
o
n
_
h
e
a
ds
12
N
um
be
r
of
a
t
t
e
n
t
i
o
n
h
e
a
ds
N
um
_
h
i
dde
n
_l
a
ye
r
s
12
N
um
be
r
of
h
i
dde
n
l
a
ye
r
s
V
oc
a
b_s
i
z
e
50265
V
oc
a
bul
a
r
y
s
i
z
e
2
.4.
3
.
A
Q
G
_C
V
_
S
q
u
a
d
m
od
e
l
A
Q
G
_C
V
_S
qua
d
m
ode
l
:
w
e
ba
s
e
t
he
m
ode
l
on
D
i
s
t
i
l
B
E
R
T
,
w
hi
c
h
i
s
a
l
i
g
ht
w
e
i
g
ht
v
e
r
s
i
on
of
B
E
R
T
t
ha
t
w
a
s
de
v
e
l
ope
d
f
or
l
ow
-
r
e
s
our
c
e
c
ont
e
xt
s
[
16]
.
I
t
pr
ov
i
de
s
f
a
s
t
e
r
s
pe
e
d
a
nd
l
i
g
ht
e
r
c
om
put
a
t
i
ona
l
bur
de
n
w
hi
c
h
m
a
ke
i
t
pos
s
i
bl
e
t
o
be
de
pl
oy
e
d
i
n
r
e
s
our
c
e
-
l
i
m
i
t
e
d
a
ppl
i
c
a
t
i
ons
.
H
ow
e
v
e
r
,
t
hi
s
opt
i
m
i
z
a
t
i
on
c
o
m
e
s
a
t
t
he
e
xpe
ns
e
of
a
s
l
i
g
ht
de
c
r
e
a
s
e
i
n a
c
c
ur
a
c
y
w
h
e
n c
o
m
p
a
r
e
d t
o i
t
s
l
a
r
ge
r
s
i
bl
i
ng
s
s
uc
h a
s
R
oB
E
R
T
a
.
S
om
e
pa
r
t
i
c
ul
a
r
de
t
a
i
l
s
of
t
he
us
e
d
m
ode
l
c
onf
i
g
ur
a
t
i
on
a
r
e
g
i
v
e
n
i
n
T
a
bl
e
4,
upon
t
he
a
r
c
hi
t
e
c
t
ur
e
f
or
da
t
a
e
xt
r
a
c
t
i
on
pr
obl
e
m
s
[
2
6
].
A
s
de
pi
c
t
e
d
i
n
T
a
bl
e
4,
t
he
A
Q
G
_C
V
_S
qua
d
m
ode
l
i
s
a
l
i
g
ht
e
r
s
upe
r
s
e
t
of
B
E
R
T
,
s
l
ow
e
r
t
o
c
o
m
put
e
but
w
i
t
h
l
e
s
s
pr
e
c
i
s
i
on,
m
a
ki
ng
i
t
a
ppr
opr
i
a
t
e
f
or
m
or
e
r
e
s
our
c
e
c
ons
t
r
a
i
ne
d
e
n
v
i
r
on
m
e
nt
.
T
a
bl
e
4. C
onf
i
g
ur
a
t
i
on pa
r
a
m
e
t
e
r
s
of
t
he
A
Q
G
_C
V
_S
qua
d m
ode
l
bui
l
t
on D
i
s
t
i
l
B
E
R
T
a
r
c
h
i
t
e
c
t
ur
e
F
e
a
t
ur
e
n
a
m
e
V
a
l
ue
D
e
s
c
r
i
pt
i
o
n
H
i
dde
n
-
act
G
e
l
u
H
i
dde
n
l
a
ye
r
a
c
t
i
va
t
i
o
n
f
u
n
c
t
i
o
n
H
i
dde
n_s
i
z
e
768
H
i
dde
n
l
a
ye
r
s
i
z
e
(
e
m
be
ddi
ng
di
m
e
n
s
i
o
n
s
)
M
a
x
_pos
i
t
i
o
n_e
m
be
ddi
ng
s
512
M
a
x
i
m
u
m
pos
i
t
i
o
n
e
m
be
ddi
ng
s
l
i
m
i
t
M
ode
l
_t
y
pe
D
i
s
t
i
l
be
r
t
M
ode
l
t
ype
N
-
h
e
a
ds
12
N
um
be
r
of
a
t
t
e
n
t
i
o
n
h
e
a
ds
N
-
l
a
ye
r
s
6
N
um
be
r
of
h
i
dde
n
l
a
ye
r
s
V
oc
a
b_s
i
z
e
30522
V
oc
a
bul
a
r
y
s
i
z
e
2
.
5
. I
n
f
or
m
a
t
i
on
e
xt
r
ac
t
i
on
u
s
i
n
g
T
S
H
D
al
gor
i
t
h
m
an
d
t
r
a
n
s
f
or
m
e
r
m
od
e
l
s
T
he
t
r
a
n
s
f
or
m
e
r
-
ba
s
e
d
m
ode
l
s
ha
v
e
r
e
v
ol
ut
i
on
i
z
e
d
t
he
f
i
e
l
d
of
N
L
P
,
a
nd
i
n
pa
r
t
i
c
u
l
a
r
i
n t
he
d
om
a
i
n
of
i
nf
or
m
a
t
i
on
e
xt
r
a
c
t
i
on
.
D
ue
t
o
t
he
i
r
c
a
pa
b
i
l
i
t
i
e
s
i
n
un
de
r
s
t
a
n
di
ng
c
on
t
e
x
t
a
n
d
s
e
m
a
nt
i
c
s
t
he
y
ha
d
m
a
j
or
i
m
pr
o
v
e
m
e
nt
s
i
n
q
ue
s
t
i
on
a
n
s
w
e
r
i
ng
,
s
um
m
a
r
i
z
a
t
i
o
n
or
doc
um
e
nt
c
l
a
s
s
i
f
i
c
a
t
i
o
n
.
T
he
s
e
t
r
a
i
t
s
,
a
n
d
t
he
i
r
pr
ope
r
t
y
a
s
g
oo
d
c
a
ndi
da
t
e
s
t
o
f
us
e
w
i
t
h
s
e
g
m
e
nt
a
t
i
on
a
ppr
oa
c
he
s
l
i
ke
T
S
H
D
t
o
i
m
pr
ov
e
t
he
e
xt
r
a
c
t
i
on
a
c
c
ur
a
c
y
.
I
n
t
hi
s
w
or
k,
w
e
pr
opos
e
a
n
i
nnov
a
t
i
v
e
m
e
t
hod
w
hi
c
h
i
nt
e
g
r
a
t
e
s
s
e
gm
e
nt
-
l
e
v
e
l
i
nf
or
m
a
t
i
on
e
xt
r
a
c
t
i
on
ba
s
e
d
on
T
S
H
D
a
l
g
or
i
t
h
m
.
I
n
c
ont
r
a
s
t
,
a
s
a
n
a
l
t
e
r
na
t
i
v
e
t
o
pr
oc
e
s
s
i
ng
t
he
e
nt
i
r
e
doc
um
e
nt
,
t
he
pr
opos
e
d
a
ppr
oa
c
h
t
a
ke
s
a
d
v
a
nt
a
g
e
of
s
e
m
a
nt
i
c
a
l
l
y
c
oh
e
r
e
nt
s
e
gm
e
nt
s
e
xt
r
a
c
t
e
d
b
y
T
S
H
D
w
hi
c
h
a
r
e
l
e
v
e
r
a
g
e
d
a
s
t
he
c
ont
e
xt
i
nput
f
or
e
xt
r
a
c
t
i
on
m
ode
l
s
.
T
hi
s
a
ppr
oa
c
h
i
m
p
r
o
v
e
s
t
he
pr
e
c
i
s
i
on
a
nd
s
a
v
e
s
c
o
m
put
a
t
i
on r
e
s
our
c
e
s
by
l
i
m
i
t
i
ng
t
he
i
nput
t
o c
ont
e
xt
ua
l
l
y
r
e
l
e
v
a
nt
s
e
gm
e
nt
s
w
i
t
h c
l
e
a
r
t
opi
c
s
.
F
i
r
s
t
, t
he
g
e
ne
r
a
l
m
e
c
h
a
ni
s
m
of
i
nf
or
m
a
t
i
on e
xt
r
a
c
t
i
on
w
i
t
h t
r
a
ns
f
or
m
e
r
-
ba
s
e
d
m
ode
l
s
i
s
i
nt
r
oduc
e
d.
I
t
t
he
n
di
s
c
us
s
e
s
a
n
e
l
a
bor
a
t
e
m
e
t
hodol
og
y
f
or
us
e
r
T
S
H
D
a
l
g
o
r
i
t
hm
i
nt
e
g
r
a
t
i
on
f
or
e
xt
r
a
c
t
i
ng
i
nf
or
m
a
t
i
on
m
o
r
e
e
f
f
e
c
t
i
v
e
l
y
.
A
s
s
how
n
i
n
F
i
g
ur
e
1
,
t
hi
s
i
s
t
he
g
e
ne
r
i
c
i
nput
/
out
pu
t
s
t
r
uc
t
ur
e
of
E
Q
A
m
od
e
l
s
,
i
nc
l
udi
ng
pr
opos
e
d s
e
gm
e
nt
a
t
i
on
-
ba
s
e
d i
m
pr
o
v
e
m
e
nt
.
A
s
s
how
n
i
n
F
i
g
ur
e
1
,
t
he
E
Q
A
m
ode
l
c
ons
i
s
t
s
of
t
hr
e
e
m
a
i
n
s
t
a
g
e
s
:
t
oke
ni
z
a
t
i
on
a
nd
i
nput
e
nc
odi
ng
,
pr
e
di
c
t
i
on
of
s
t
a
r
t
a
nd
e
nd
l
o
g
i
t
s
,
a
nd
e
xt
r
a
c
t
i
on
of
t
he
f
i
na
l
a
ns
w
e
r
s
pa
n.
I
n
our
m
e
t
hod
t
he
c
ont
e
xt
i
s
c
onf
i
ne
d
t
o
s
e
m
a
nt
i
c
a
l
l
y
r
e
l
e
v
a
nt
pa
r
t
s
,
a
s
a
ddr
e
s
s
e
d
by
t
he
T
S
H
D
a
l
g
or
i
t
h
m
.
F
i
g
ur
e
1
,
a
l
s
o
s
how
s
how
t
o us
e
t
he
m
ode
l
i
n i
nf
or
m
a
t
i
on e
xt
r
a
c
t
i
on t
o t
he
e
xt
r
a
c
t
que
s
t
i
on (
t
he
i
nf
or
m
a
t
i
on i
n c
ont
e
xt
)
;
t
he
m
od
e
l
f
ol
l
ow
s
t
he
ne
xt
s
t
e
ps
t
o
i
de
n
t
i
f
y
a
nd r
e
t
ur
n t
he
a
ns
w
e
r
f
r
o
m
t
he
pr
o
v
i
d
e
d t
e
xt
:
a)
P
r
e
-
pr
oc
e
s
s
i
ng
:
t
he
i
nput
da
t
a
i
s
pr
e
-
pr
oc
e
s
s
e
d,
s
o
i
t
’
s
s
ui
t
a
bl
e
f
or
t
he
m
ode
l
.
T
hi
s
m
e
a
n
s
pe
r
f
or
m
i
n
g
t
oke
ni
z
a
t
i
on
a
nd
t
r
a
ns
f
or
m
i
ng
t
he
t
e
xt
ua
l
i
nput
i
n
t
o
t
he
a
ppr
opr
i
a
t
e
I
D
s
i
n
t
he
l
e
xi
c
on
of
t
he
m
ode
l
’
s
v
oc
a
bul
a
r
y
,
t
he
n r
unni
ng
t
he
da
t
a
t
hr
oug
h t
he
m
ode
l
.
b)
M
ode
l
pr
e
di
c
t
i
on:
t
he
m
od
e
l
pr
e
di
c
t
s
t
he
s
t
a
r
t
l
o
g
i
t
a
nd
e
nd
l
og
i
t
f
or
e
a
c
h
t
oke
n
(
w
he
r
e
a
t
oke
n
i
s
a
l
a
ng
ua
g
e
uni
t
)
,
s
pe
c
i
f
y
i
n
g
how
l
i
ke
l
y
(
w
i
t
h
r
e
s
pe
c
t
t
o
t
he
t
e
xt
ua
l
c
ont
e
xt
i
n
w
hi
c
h
t
he
y
a
r
e
pr
e
s
e
nt
)
i
s
t
ha
t
t
oke
n t
o be
t
he
s
t
a
r
t
/
e
nd of
t
he
a
ns
w
e
r
.
c)
P
os
t
-
pr
oc
e
s
s
i
ng
:
d
ur
i
ng
t
hi
s
f
i
na
l
s
t
a
g
e
,
t
he
m
ode
l
e
xa
m
i
n
e
s
t
he
l
og
i
t
s
i
t
g
e
ne
r
a
t
e
d
t
o
f
i
nd
t
he
t
oke
ns
w
i
t
h
t
he
hi
g
he
s
t
pr
oba
bi
l
i
t
y
,
t
hus
de
t
e
r
m
i
ni
ng
w
h
e
r
e
t
he
a
ns
w
e
r
s
t
a
r
t
s
a
nd
e
nds
i
n
t
he
t
e
x
t
ua
l
c
ont
e
xt
.
T
hi
s
s
ki
l
l
pr
ov
i
de
s
a
n a
c
c
ur
a
t
e
a
ns
w
e
r
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2502
-
4752
I
ndone
s
i
a
n J
E
l
e
c
E
n
g
&
C
om
p S
c
i
, V
ol
.
40
, N
o.
3
,
D
e
c
e
m
be
r
20
25
:
1
453
-
1
46
5
1460
N
e
xt
,
w
e
pr
e
s
e
nt
how
t
he
i
nf
or
m
a
t
i
on
e
xt
r
a
c
t
i
on
m
e
c
h
a
ni
s
m
w
i
t
h
T
S
H
D
a
l
g
o
r
i
t
hm
i
s
c
om
bi
ne
d
.
I
ns
t
e
a
d
of
s
e
ndi
ng
t
he
f
ul
l
doc
um
e
nt
t
o
be
pr
oc
e
s
s
e
d,
onl
y
t
he
e
xt
r
a
c
t
e
d
s
e
gm
e
nt
s
a
l
ong
w
i
t
h
t
he
i
r
c
or
r
e
s
pondi
ng
t
opi
c
s
a
r
e
s
e
nt
t
o
t
he
m
ode
l
f
or
pr
oc
e
s
s
i
ng
.
T
hi
s
f
oc
us
e
d
m
e
t
hod
m
a
xi
m
i
z
e
s
a
c
c
ur
a
c
y
a
n
d
e
f
f
i
c
i
e
nc
y
. F
i
g
ur
e
2
i
l
l
us
t
r
a
t
e
s
how
t
he
m
ode
l
c
o
m
bi
ne
s
f
ul
l
r
e
s
u
m
e
s
e
gm
e
nt
a
t
i
on by
T
S
H
D
, s
e
c
t
i
on s
e
l
e
c
t
i
on
a
c
c
or
di
ng
t
o
t
he
a
l
i
g
n
m
e
nt
w
i
t
h
t
he
que
r
y
t
opi
c
,
a
nd
f
i
na
l
l
y
Q
A
m
ode
l
t
o
e
xt
r
a
c
t
t
he
e
x
a
c
t
a
ns
w
e
r
s
f
r
o
m
t
he
f
i
l
t
e
r
e
d c
ont
e
xt
.
I
n
t
he
f
i
r
s
t
s
t
e
p,
t
he
f
ul
l
doc
u
m
e
nt
i
s
f
e
d
i
nt
o
t
he
T
S
H
D
m
ode
l
,
w
hi
c
h
s
e
gm
e
nt
s
e
a
c
h
pa
s
s
a
g
e
a
nd
i
de
nt
i
f
i
e
s
i
t
s
t
opi
c
.
A
f
t
e
r
t
ha
t
,
t
he
que
s
t
i
on
i
s
m
a
t
c
h
e
d
w
i
t
h
t
he
r
e
l
e
v
a
nt
s
e
gm
e
nt
ba
s
e
d
on
t
opi
c
.
T
he
a
ppr
o
a
c
h
g
i
v
e
s
t
he
m
ode
l
onl
y
t
he
que
s
t
i
on,
a
nd
t
he
r
e
l
e
v
a
nt
pa
r
t
of
t
he
doc
um
e
nt
s
o
i
t
doe
s
not
ha
v
e
t
o
de
a
l
w
i
t
h
t
he
e
nt
i
r
e
doc
u
m
e
nt
.
T
he
n,
ba
s
e
d
on
t
he
r
e
t
r
i
e
v
e
d
s
ni
ppe
t
,
t
he
m
od
e
l
e
xt
r
a
c
t
s
r
e
l
e
v
a
nt
i
nf
or
m
a
t
i
on
(
i
.e
.
,
t
he
a
ns
w
e
r
)
,
s
a
v
e
s
t
he
r
e
s
ul
t
i
n
a
s
t
r
uc
t
ur
e
d
f
or
m
a
t
.
B
y
honi
ng
i
n
on
c
ont
e
xt
ua
l
l
y
r
e
l
e
v
a
nt
s
e
gm
e
nt
s
,
t
hi
s
m
e
t
hod
i
m
pr
o
v
e
s
t
he
pr
e
c
i
s
i
on
a
nd
e
f
f
i
c
i
e
nc
y
of
i
nf
or
m
a
t
i
on
r
e
t
r
i
e
v
a
l
,
a
s
oppos
e
d
t
o
pa
r
s
i
ng
t
he
f
ul
l
doc
u
m
e
nt
. T
h
e
e
xpe
r
i
m
e
nt
a
l
s
e
t
up w
e
us
e
f
ol
l
ow
s
t
he
s
t
e
ps
:
a)
I
n
t
hi
s
w
or
k,
w
e
c
ol
l
e
c
t
e
d
105
E
ng
l
i
s
h
r
e
s
u
m
e
s
of
di
f
f
e
r
e
nt
hi
s
t
or
y
f
r
o
m
onl
i
ne
por
t
f
ol
i
o,
a
nd
m
a
nu
a
l
l
y
l
a
be
l
e
d
t
he
f
i
e
l
ds
(
na
m
e
,
e
m
a
i
l
,
phone
nu
m
be
r
,
a
nd
e
du
c
a
t
i
on)
i
n
t
he
r
e
s
u
m
e
s
t
o
obt
a
i
n
a
l
a
be
l
e
d
da
t
a
s
e
t
, a
nd w
e
a
dopt
e
d t
he
S
Q
uA
D
1.1
w
a
y
t
o a
r
r
a
n
g
e
t
he
que
s
t
i
ons
a
nd t
he
a
ns
w
e
r
s
.
b)
P
r
e
-
p
r
oc
e
s
s
i
ng
w
i
t
h
T
S
H
D
a
l
g
or
i
t
h
m
:
T
he
r
e
s
u
m
e
s
w
e
r
e
pr
e
-
pr
oc
e
s
s
e
d
w
i
t
h
t
he
T
S
H
D
a
l
g
or
i
t
h
m
t
o
e
xt
r
a
c
t
ke
y
-
v
a
l
ue
pa
i
r
c
ont
e
nt
s
t
or
e
d
i
n
j
s
on
f
or
m
a
t
,
w
he
r
e
e
a
c
h
ke
y
i
s
t
he
s
e
c
t
i
on
he
a
de
r
e
xt
r
a
c
t
e
d
a
n
d
w
a
s
a
s
s
i
g
ne
d a
v
a
l
ue
w
hi
c
h i
s
t
he
t
e
xt
of
t
he
c
ont
e
nt
f
ound f
or
t
ha
t
s
e
c
t
i
on.
c)
I
n
m
ode
l
s
e
l
e
c
t
i
on
a
nd
c
onf
i
g
ur
a
t
i
on,
t
hr
e
e
t
r
a
ns
f
or
m
e
r
ba
s
e
d
m
ode
l
s
i
.e
.
,
a
ut
ot
r
a
i
n
-
r
e
s
u
m
e
_pa
r
s
e
r
(
L
on
g
f
or
m
e
r
)
,
C
V
_C
us
t
om
_D
S
(
R
oB
E
R
T
a
)
a
nd
A
Q
G
_C
V
_S
qua
d(
D
i
s
t
i
l
B
E
R
T
)
w
e
r
e
c
hos
e
n, a
nd a
l
l
m
ode
l
s
w
e
r
e
f
i
ne
-
t
une
d on t
he
pr
oc
e
s
s
e
d da
t
a
s
e
t
us
i
ng
P
yT
o
r
c
h.
d)
I
n
B
a
s
e
l
i
ne
w
i
t
hout
T
S
H
D
,
or
i
g
i
na
l
r
e
s
u
m
e
s
w
e
r
e
e
v
a
l
ua
t
e
d
di
r
e
c
t
l
y
w
i
t
hout
be
i
ng
s
e
gm
e
nt
e
d,
a
n
d
m
od
e
l
s
’
pr
e
di
c
t
i
on w
e
r
e
c
he
c
ke
d a
g
a
i
ns
t
t
he
g
r
ound t
r
ut
h us
i
ng
t
he
m
e
t
r
i
c
s
EM
a
nd
F1
-
s
c
or
e
.
e)
E
v
a
l
ua
t
i
on
w
i
t
h
T
S
H
D
I
nt
e
g
r
a
t
i
on
F
or
e
a
c
h
que
s
t
i
on,
t
he
T
S
H
D
-
s
e
gm
e
nt
e
d
r
e
s
u
m
e
w
a
s
a
na
l
y
s
e
d
t
o
de
t
e
r
m
i
ne
w
hi
c
h
s
e
c
t
i
on
i
s
m
os
t
r
e
l
e
v
a
nt
t
o
r
e
s
um
e
c
ont
e
nt
by
t
opi
c
m
a
t
c
hi
ng
a
nd
t
he
s
e
l
e
c
t
e
d
por
t
i
on
of
t
he
r
e
s
u
m
e
s
e
r
v
e
d
a
s
t
he
i
nput
i
ns
t
e
a
d
of
t
he
w
hol
e
doc
u
m
e
nt
t
o
t
he
Q
A
m
ode
l
,
w
i
t
h
r
e
s
ul
t
i
ng
a
ns
w
e
r
s
a
g
a
i
n e
v
a
l
ua
t
e
d us
i
ng
E
M
a
nd
F1
-
s
c
or
e
s.
f)
P
e
r
f
or
m
a
n
c
e
c
o
m
p
a
r
i
s
on
i
nv
ol
v
e
d
t
he
c
a
l
c
ul
a
t
i
on
a
nd
s
t
a
t
i
s
t
i
c
a
l
a
n
a
l
y
s
i
s
of
t
he
di
f
f
e
r
e
nc
e
s
be
t
w
e
e
n
ba
s
e
l
i
ne
a
nd L
e
ukon
y
c
hi
a
T
S
H
D
-
e
nha
nc
e
d r
e
s
ul
t
s
.
T
he
m
e
t
hod
a
s
de
s
c
r
i
be
d
a
bo
v
e
e
ns
ur
e
s
t
he
f
ul
l
r
e
pr
oduc
i
bi
l
i
t
y
,
a
nd
t
he
de
v
e
l
ope
d
m
o
de
l
s
a
nd
da
t
a
s
e
t
s
us
e
d a
r
e
publ
i
c
l
y a
v
a
i
l
a
bl
e
,
w
hi
c
h
m
a
ke
s
i
t
pos
s
i
bl
e
t
o ha
v
e
r
e
p
r
oduc
i
bl
e
, t
r
a
ns
pa
r
e
nt
a
nd
v
a
l
i
da
t
i
on.
I
n
t
he
ne
xt
s
e
c
t
i
on,
w
e
e
xa
m
i
ne
t
he
pe
r
f
or
m
a
nc
e
of
our
pr
opos
e
d
m
e
t
hod
a
nd
t
he
c
om
pa
r
i
s
on
be
t
w
e
e
n
t
he
pe
r
f
or
m
a
nc
e
of
t
he
m
ode
l
on T
S
H
D
w
i
t
h a
nd w
i
t
hout
s
e
gm
e
nt
a
t
i
on.
F
i
g
ur
e
1
. I
nput
-
out
put
s
t
r
uc
t
ur
e
of
E
Q
A
m
od
e
l
s
w
i
t
h
pr
oc
e
s
s
i
ng
s
t
a
g
e
s
F
i
g
ur
e
2
. W
or
kf
l
ow
of
i
nt
e
g
r
a
t
i
ng
t
he
T
S
H
D
a
l
g
or
i
t
h
m
w
i
t
h
t
r
a
ns
f
or
m
e
r
-
ba
s
e
d e
xt
r
a
c
t
i
v
e
Q
A
m
ode
l
s
3.
R
E
S
U
L
T
S
A
N
D
D
I
S
C
U
S
S
I
O
N
I
n
or
de
r
t
o
a
na
l
y
s
e
t
he
c
o
m
pl
e
m
e
nt
a
r
i
t
y
of
our
a
ppr
oa
c
h,
w
e
e
m
pi
r
i
c
a
l
l
y
e
v
a
l
ua
t
e
d
t
he
pe
r
f
om
a
nc
e
s
w
i
t
h
3
m
ode
l
s
ba
s
e
d
on
t
r
a
ns
f
or
m
e
r
t
r
a
i
ne
d
on
r
e
s
u
m
e
c
or
pus
.
S
e
l
e
c
t
e
d
e
v
a
l
ua
t
i
on
r
e
s
ul
t
s
of
t
he
pr
opos
e
d
m
e
t
hod
a
r
e
de
m
ons
t
r
a
t
e
d,
by
f
oc
us
i
ng
on
t
he
m
a
i
n
m
e
t
r
i
c
s
,
i
nc
l
udi
ng
EM
a
nd
F1
-
s
c
o
r
e
s
i
g
na
l
s
.
I
n
t
he
e
xpe
r
i
m
e
nt
a
l
s
e
t
t
i
ng
w
e
c
o
m
pa
r
e
t
he
m
ode
l
s
w
i
t
h
a
nd
w
i
t
hout
T
S
H
D
s
e
gm
e
nt
a
t
i
on
t
o
de
m
ons
t
r
a
t
e
t
he
ga
i
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
ndone
s
i
a
n J
E
l
e
c
E
n
g
&
C
om
p S
c
i
I
S
S
N
:
2502
-
4752
O
pt
i
m
i
z
i
n
g r
e
s
um
e
i
nf
or
m
a
t
i
on e
x
t
r
ac
t
i
on t
hr
ough T
SH
D
s
e
gm
e
nt
at
i
on
an
d
…
(
A
nm
ar
A
b
u
ham
d
a
h
)
1461
br
oug
ht
by
t
opi
c
-
de
pe
nde
nt
pr
oc
e
s
s
.
A
de
m
ons
t
r
a
t
i
on
of
t
he
pr
opos
e
d
m
e
t
hod
e
v
a
l
ua
t
i
ons
i
s
s
how
n
ne
xt
,
t
he
pr
opos
e
d
m
e
t
hod
w
a
s
t
e
s
t
e
d
on t
op of
a
s
e
t
of
i
nf
or
m
a
t
i
on e
xt
r
a
c
t
i
on
m
ode
l
s
a
s
d
e
t
a
i
l
e
d on t
he
r
e
s
e
a
r
c
h
-
t
ool
s
s
e
c
t
i
on
.
W
e
pe
r
f
or
m
e
d
t
he
s
e
m
ode
l
s
i
n
t
he
r
e
s
um
e
dom
a
i
n
ba
s
e
d on t
he
S
Q
uA
D
1
.1
f
r
a
m
e
w
or
k t
o
m
e
a
s
ur
e
t
he
e
f
f
e
c
t
s
of
a
pp
l
y
i
ng
t
he
T
S
H
D
t
o
i
m
pr
ov
e
t
he
a
c
c
ur
a
c
y
ov
e
r
t
w
o
pha
s
e
s
of
i
nf
or
m
a
t
i
on
e
x
t
r
a
c
t
i
o
n.
I
nf
or
m
a
t
i
on
e
xt
r
a
c
t
i
o
n
w
a
s
e
v
a
l
ua
t
e
d
o
n
a
da
t
a
s
e
t
c
o
nt
a
i
n
i
ng
44
4
q
ue
s
t
i
o
ns
,
u
t
i
l
i
z
i
ng
m
ode
l
s
s
uc
h
a
s
C
V
_
C
us
t
om
_D
S
,
a
ut
o
t
r
a
i
n
-
r
e
s
um
e
_pa
r
s
e
r
, a
nd A
Q
G
_C
V
_S
qua
d.
M
o
de
l
e
v
a
l
ua
t
i
o
n w
a
s
pe
r
f
or
m
e
d
t
w
i
c
e
f
or
e
a
c
h one
:
W
i
t
hout
us
i
ng
T
S
H
D
a
l
g
o
r
i
t
h
m
:
h
e
r
e
, t
he
m
ode
l
i
s
f
e
t
c
hi
n
g
w
ha
t
i
t
ne
e
ds
f
r
o
m
t
he
w
hol
e
d
oc
u
m
e
nt
.
U
s
i
ng
t
he
T
S
H
D
a
l
g
or
i
t
h
m
:
i
n t
hi
s
s
t
e
p
t
he
m
ode
l
pr
o
v
i
de
s
t
he
ne
c
e
s
s
a
r
y
i
nf
or
m
a
t
i
on e
xt
r
a
c
t
e
d
f
r
o
m
t
h
e
r
e
l
a
t
i
v
e
s
e
gm
e
nt
ba
s
e
d on t
he
de
t
e
r
m
i
n
e
d t
opi
c
.
E
v
a
l
ua
t
i
on
r
e
s
ul
t
s
of
a
l
l
t
hr
e
e
m
ode
l
s
i
n
t
e
r
m
s
of
i
nf
or
m
a
t
i
on
e
xt
r
a
c
t
i
on
on
t
he
t
e
s
t
da
t
a
s
e
t
w
i
t
h
a
nd
w
i
t
ho
ut
T
S
H
D
a
l
g
or
i
t
hm
a
r
e
s
how
n
i
n
T
a
bl
e
1
,
a
nd
a
l
s
o
i
l
l
us
t
r
a
t
e
d i
n F
i
g
ur
e
s
3
a
n
d
4
.
A
s
r
e
por
t
e
d
i
n
T
a
bl
e
1,
T
S
H
D
s
e
gm
e
nt
a
t
i
on
not
i
c
e
a
bl
y
i
nc
r
e
a
s
e
s
bot
h
E
M
a
nd
F1
-
s
c
or
e
s
f
or
a
l
l
m
ode
l
s
,
w
hi
c
h
i
ndi
c
a
t
e
s
t
he
e
f
f
e
c
t
i
v
e
ne
s
s
of
t
opi
c
-
a
w
a
r
e
c
ont
e
xt
f
i
l
t
e
r
i
ng
.
F
i
g
ur
e
3
s
how
s
a
g
r
a
phi
c
a
l
s
u
m
m
a
r
y
of
E
M
s
c
or
e
s
of
a
l
l
t
hr
e
e
m
ode
l
s
be
f
or
e
a
nd
a
f
t
e
r
t
h
e
T
S
H
D
-
ba
s
e
d s
e
gm
e
nt
a
t
i
on.
T
he
f
i
ndi
ng
s
f
ur
t
he
r
de
m
ons
t
r
a
t
e
t
ha
t
f
o
c
us
i
ng
on t
he
r
e
l
e
v
a
nt
s
e
c
t
i
ons
of
t
he
doc
u
m
e
nt
s
i
m
pr
o
v
e
s
t
he
E
M
s
c
or
e
s
uni
f
or
m
l
y
, i
n pa
r
t
i
c
ul
a
r
f
or
R
oB
E
R
T
a
a
nd D
i
s
t
i
l
B
E
R
T
, f
or
r
e
t
r
i
e
v
a
l
of
a
ns
w
e
r
s
.
T
a
bl
e
1. P
e
r
f
or
m
a
nc
e
c
o
m
p
a
r
i
s
on of
t
hr
e
e
m
od
e
l
s
w
i
t
h a
nd w
i
t
hou
t
T
S
H
D
s
e
gm
e
nt
a
t
i
on
M
ode
l
W
i
t
hout
T
S
H
D
a
l
g
o
r
i
t
hm
F1
-
s
c
or
e
(%)
W
i
t
h
T
S
H
D
a
l
g
or
i
t
hm
E
M
(
%
)
A
ut
ot
r
a
i
n
-
r
e
s
u
m
e
_pa
r
s
e
r
91.
55
80.
63
C
V
_
c
us
t
o
m
_D
S
73.
38
68.
69
A
Q
G
_C
V
_
s
qua
d
56.
23
40.
54
F
r
om
F
i
g
ur
e
4
,
w
h
e
n
T
S
H
D
w
a
s
i
nt
r
oduc
e
d
t
o
t
he
pr
oc
e
s
s
i
ng
pi
pe
l
i
ne
,
a
l
l
t
hr
e
e
m
ode
l
s
’
F1
-
s
c
or
e
s
i
m
pr
o
v
e
d
dr
a
m
a
t
i
c
a
l
l
y
.
S
i
g
ni
f
i
c
a
nt
l
y
,
t
he
R
oB
E
R
T
a
-
ba
s
e
d
m
ode
l
s
a
w
t
he
g
r
e
a
t
e
s
t
g
a
i
n,
s
uppor
t
i
ng
t
he
e
f
f
e
c
t
i
v
e
ne
s
s
of
t
he
t
opi
c
-
a
w
a
r
e
s
e
gm
e
nt
a
t
i
on i
n de
c
r
e
a
s
i
ng
noi
s
e
a
nd i
nc
r
e
a
s
i
ng
pr
e
c
i
s
i
on.
A
s
s
how
n
i
n
F
i
g
ur
e
s
3
a
nd
4
a
nd
T
a
bl
e
1,
t
he
us
e
d
of
t
he
T
S
H
D
a
l
g
or
i
t
h
m
s
i
g
ni
f
i
c
a
nt
l
y
i
m
p
r
o
v
e
d
t
he
e
v
a
l
ua
t
i
on r
e
s
ul
t
s
of
t
he
t
hr
e
e
di
f
f
e
r
e
nt
m
od
e
l
s
on di
f
f
e
r
e
nt
m
e
t
r
i
c
s
. S
pe
c
i
f
i
c
a
l
l
y
:
a)
F
or
t
he
a
ut
ot
r
a
i
n
-
r
e
s
um
e
_pa
r
s
e
r
m
ode
l
:
T
he
E
M
m
e
t
r
i
c
w
e
nt
f
r
o
m
80.63%
(
358
a
ns
w
e
r
s
)
up
t
o
83.56%
(
371
a
ns
w
e
r
s
)
,
a
2.93%
(
13
a
ns
w
e
r
s
)
i
m
pr
o
v
e
m
e
nt
.
T
he
F1
-
s
c
or
e
a
l
s
o
i
nc
r
e
a
s
e
d f
r
o
m
91.55%
t
o 93.18%
, a
n i
m
pr
o
v
e
m
e
nt
of
1.63%
b)
M
ode
l
C
V
_
C
us
t
om
_D
S
E
M
m
e
t
r
i
c
:
68.69%
(
305 a
ns
w
e
r
s
)
=
>
76.13%
(
338 a
ns
w
e
r
s
)
+
7.44%
(
+
33 a
ns
w
e
r
s
)
.
T
he
F
1
-
s
c
or
e
i
nc
r
e
a
s
e
d f
r
o
m
73.38%
t
o 81.12%
w
i
t
h a
g
a
i
n of
7.74%
c)
F
or
A
Q
G
_C
V
_S
qua
d m
ode
l
E
M
m
e
t
r
i
c
i
nc
r
e
a
s
e
d
f
r
o
m
40.54%
(
180
a
ns
w
e
r
s
)
t
o
45.95%
(
204
a
ns
w
e
r
s
)
,
a
5.41%
(
24
a
ns
w
e
r
s
)
i
m
pr
o
v
e
m
e
nt
.
T
he
F1
-
s
c
or
e
m
e
t
r
i
c
w
a
s
i
nc
r
e
a
s
e
d
f
r
o
m
56.23%
t
o 64.07%
, a
pr
e
c
i
s
e
ga
i
n of
7.84%
.
O
v
e
r
a
l
l
, t
hi
s
s
i
g
ni
f
i
c
a
nt
l
y
e
nha
n
c
e
d t
he
i
nf
or
m
a
t
i
on r
e
t
r
i
e
v
a
l
a
c
c
ur
a
c
y of
t
r
a
ns
f
or
m
e
r
m
od
e
l
s
w
he
n
c
o
m
bi
ne
d
w
i
t
h t
he
t
he
T
S
H
D
a
l
g
or
i
t
h
m
.
EM
m
e
t
r
i
c
i
nc
r
e
a
s
e
d f
r
o
m
2.93%
t
o 7.44%
a
nd
F1
-
s
c
or
e
m
e
t
r
i
c
i
nc
r
e
a
s
e
d
f
r
o
m
1.63%
t
o 7.84%
.
D
i
s
c
r
e
pa
nc
i
e
s
be
t
w
e
e
n
t
he
pe
r
f
or
m
a
nc
e
s
of
t
he
m
ode
l
s
a
r
e
r
e
l
a
t
e
d
t
o
t
he
i
r
m
ode
l
a
r
c
hi
t
e
c
t
ur
e
s
a
nd
l
e
a
r
ni
ng
m
e
t
hodol
og
i
e
s
[
11]
.
T
hi
s
i
s
f
or
i
ns
t
a
nc
e
t
he
m
ode
l
A
Q
G
_C
V
_S
qua
d, ba
s
e
d on D
i
s
t
i
l
B
E
R
T
, t
ha
t
ha
d
t
he
l
ow
e
s
t
pe
r
f
or
m
a
nc
e
of
t
he
m
ode
l
s
t
e
s
t
e
d
[
21]
,
[
27
]
.
T
hi
s
r
e
s
ul
t
i
s
c
ons
i
s
t
e
nt
w
i
t
h
pr
i
or
w
o
r
k
i
n
w
hi
c
h
s
m
a
l
l
e
r
m
ode
l
s
m
a
ke
a
n
a
c
c
ur
a
c
y
-
c
o
m
put
a
t
i
ona
l
c
os
t
t
r
a
de
of
f
,
e
s
pe
c
i
a
l
l
y
unde
r
he
a
v
y
t
a
s
ks
s
uc
h
a
s
E
Q
A
[
20]
.
T
hi
s
s
a
c
r
i
f
i
c
e
,
a
s
T
a
bl
e
1
s
ho
w
s
,
i
s
j
us
t
i
f
i
e
d
i
n
r
e
s
u
m
e
pa
r
s
i
ng
t
a
s
ks
b
y
t
he
l
o
w
e
r
F
1
a
nd
E
M
s
c
or
e
s
.
T
hi
s
i
s
i
n
a
c
c
or
da
nc
e
t
o
our
e
xpe
c
t
a
t
i
on,
s
i
nc
e
D
i
s
t
i
l
B
E
R
T
i
s
a
s
m
a
l
l
v
e
r
s
i
on
of
B
E
R
T
t
a
i
l
or
e
d
t
o
be
c
o
m
put
a
t
i
ona
l
l
y
e
f
f
i
c
i
e
nt
but
not
a
c
c
ur
a
t
e
t
o
t
he
be
s
t
[
22]
,
[
11
]
.
I
t
s
s
m
a
l
l
e
r
s
i
z
e
a
nd
l
ow
e
r
nu
m
be
r
of
pa
r
a
m
e
t
e
r
s
m
a
ke
i
t
s
ui
t
a
bl
e
f
or
r
unni
ng
on
r
e
s
our
c
e
c
ons
t
r
a
i
ne
d
de
v
i
c
e
s
e
v
e
n
w
he
n
i
t
i
s
a
t
t
he
s
a
m
e
t
i
m
e
s
l
i
g
ht
l
y
l
ow
e
r
pe
r
f
or
m
i
n
g
on
m
o
r
e
c
o
m
pl
e
x
N
L
P
t
a
s
ks
l
i
ke
E
Q
A
.
P
r
e
v
i
ous
w
o
r
ks
ha
v
e
v
e
r
i
f
i
e
d
t
ha
t
t
he
s
e
c
o
m
pa
c
t
m
ode
l
s
c
a
n
be
de
pl
oye
d
w
i
t
h
l
i
m
i
t
e
d
r
e
s
our
c
e
s
;
how
e
v
e
r
,
t
he
i
r
pe
r
f
or
m
a
nc
e
de
g
r
a
d
e
s
w
he
n
t
he
y
pr
oc
e
s
s
m
ul
t
i
-
se
gm
e
nt
or
l
ong
-
doc
u
m
e
nt
t
a
s
ks
[
22
]
.
I
t
i
s
s
ui
t
e
d
f
or
r
u
nni
ng
i
n
r
e
s
our
c
e
-
c
ons
t
r
a
i
ne
d
e
n
v
i
r
on
m
e
nt
s
due
t
o
i
t
s
s
m
a
l
l
e
r
s
i
z
e
a
nd
f
e
w
e
r
pa
r
a
m
e
t
e
r
s
,
how
e
v
e
r
w
i
t
h
a
t
r
a
de
-
of
f
be
i
ng
a
l
ow
e
r
pe
r
f
or
m
a
nc
e
i
n
c
o
m
pl
e
x
N
L
P
t
a
s
ks
s
uc
h
a
s
E
Q
A
[
23
]
.
T
hi
s
i
s
a
n
e
xa
m
pl
e
of
t
he
-
t
r
a
de
of
f
be
t
w
e
e
n
t
he
t
a
s
k
de
m
a
nd
a
nd
t
he
a
v
a
i
l
a
bl
e
r
e
s
our
c
e
,
a
nd
t
ha
t
t
he
c
ho
i
c
e
of
t
he
m
ode
l
t
o
be
us
e
d
c
ou
l
d
be
de
t
e
r
m
i
ne
d
a
c
c
or
di
ng
t
o
t
he
bot
h
t
a
s
k
de
m
a
nd
a
nd
t
he
a
v
a
i
l
a
bl
e
r
e
s
our
c
e
.
T
h
e
a
d
v
a
nt
a
ge
of
doi
ng
t
hi
s
i
s
t
ha
t
i
t
i
s
f
a
s
t
e
r
a
nd r
e
qui
r
e
s
l
e
s
s
c
o
m
put
a
t
i
on, t
houg
h l
e
s
s
a
c
c
ur
a
t
e
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2502
-
4752
I
ndone
s
i
a
n J
E
l
e
c
E
n
g
&
C
om
p S
c
i
, V
ol
.
40
, N
o.
3
,
D
e
c
e
m
be
r
20
25
:
1
453
-
1
46
5
1462
F
i
g
ur
e
3
. C
om
pa
r
a
t
i
v
e
a
na
l
y
s
i
s
of
E
M
s
c
or
e
s
a
c
r
os
s
t
hr
e
e
t
r
a
ns
f
or
m
e
r
-
ba
s
e
d
m
o
d
e
l
s
w
i
t
h a
nd
w
i
t
hout
T
S
H
D
s
e
gm
e
nt
a
t
i
on
F
i
g
ur
e
4
. C
om
pa
r
a
t
i
v
e
a
na
l
y
s
i
s
of
F1
-
s
c
or
e
s
a
c
r
os
s
t
hr
e
e
t
r
a
ns
f
or
m
e
r
-
ba
s
e
d
m
od
e
l
s
w
i
t
h a
n
d
w
i
t
hout
T
S
H
D
s
e
gm
e
nt
a
t
i
on
4.
C
O
N
C
L
U
S
I
O
N
T
he
pr
opos
e
d
r
e
s
e
a
r
c
h
pr
opos
e
s
a
ne
w
s
e
gm
e
nt
a
t
i
on
m
e
t
hod
ba
s
e
d
on
T
S
H
D
t
o
i
m
p
r
o
v
e
t
h
e
i
nf
or
m
a
t
i
on
e
xt
r
a
c
t
i
on
m
e
c
ha
ni
s
m
of
t
r
a
ns
f
or
m
e
r
m
ode
l
s
.
T
he
a
ppr
oa
c
h
i
s
t
o
r
un
t
he
s
a
m
e
a
l
g
or
i
t
h
m
t
o
e
xt
r
a
c
t
bl
oc
ks
of
t
e
xt
,
i
de
nt
i
f
y
t
he
t
opi
c
f
or
t
he
s
e
bl
oc
ks
of
t
e
xt
a
nd
t
he
n
a
l
l
ow
t
he
m
od
e
l
s
t
o
e
xt
r
a
c
t
i
nf
or
m
a
t
i
on f
r
o
m
t
he
s
e
bl
oc
ks
i
ns
t
e
a
d of
t
he
w
hol
e
doc
u
m
e
nt
.
T
hr
e
e
m
ode
l
s
w
e
r
e
t
e
s
t
e
d
a
nd
e
v
a
l
ua
t
e
d
on
a
c
ol
l
e
c
t
i
on
of
r
e
s
um
e
s
w
i
t
h
a
nd
w
i
t
hout
T
S
H
D
,
r
e
pl
i
c
a
t
i
ng
t
he
unpr
e
c
e
d
e
nt
e
d
s
t
r
uc
t
ur
e
o
f
t
he
S
Q
u
A
D
1.1
d
a
t
a
s
e
t
.
E
xpe
r
i
m
e
nt
a
l
r
e
s
ul
t
s
s
how
a
n
ob
v
i
ou
s
i
m
pr
o
v
e
m
e
nt
a
f
t
e
r
us
i
n
g
t
he
T
S
H
D
a
l
g
or
i
t
h
m
, a
c
hi
e
v
i
n
g
a
m
a
xi
m
a
l
i
nc
r
e
a
s
e
of
7.44%
on t
he
EM
m
e
t
r
i
c
,
a
nd
a
m
a
xi
m
a
l
i
nc
r
e
a
s
e
o
f
7.8%
on
t
he
F1
-
s
c
or
e
m
e
t
r
i
c
.
M
or
e
o
v
e
r
,
T
S
H
D
a
l
g
o
r
i
t
h
m
s
t
i
l
l
c
ont
r
i
but
e
s
t
o
t
he
s
m
a
l
l
e
r
a
m
ount
of
c
o
m
p
ut
a
t
i
ona
l
r
e
s
our
c
e
s
ne
e
de
d
be
c
a
us
e
of
t
he
de
c
r
e
a
s
e
d
di
m
e
ns
i
ona
l
i
t
y
of
t
he
t
e
xt
ua
l
c
ont
e
xt
.
A
ddi
t
i
ona
l
l
y
,
i
t
s
v
i
s
a
ge
doe
s
not
de
m
a
nd
r
e
t
r
a
i
ni
ng
t
he
m
ode
l
s
on
t
he
doc
u
m
e
nt
s
e
gm
e
nt
s
a
f
t
e
r
s
e
gm
e
nt
a
t
i
on.
T
he
i
m
pl
i
c
a
t
i
ons
i
n
t
e
r
m
s
of
a
ut
o
m
a
t
e
d
r
e
c
r
ui
t
m
e
nt
s
y
s
t
e
m
s
a
r
e
c
l
e
a
r
:
q
ui
c
k
a
nd
a
c
c
ur
a
t
e
s
e
l
e
c
t
i
on
of
c
a
ndi
da
t
e
s
i
s
of
g
r
e
a
t
i
m
po
r
t
a
nc
e
.
O
u
r
m
e
t
hod
e
na
bl
e
s
s
c
a
l
a
bl
e
H
R
a
ut
o
m
a
t
i
on
a
nd
e
nha
nc
e
d
de
c
i
s
i
on
-
m
a
ki
ng
pi
pe
l
i
ne
s
by
i
m
pr
o
v
i
n
g
pr
e
c
i
s
i
on a
nd r
e
duc
i
n
g
c
o
m
put
a
t
i
ona
l
bur
de
ns
.
O
ur
f
i
ndi
ng
s
s
ug
g
e
s
t
t
ha
t
c
o
m
bi
ni
ng
t
opi
c
-
a
w
a
r
e
s
e
gm
e
nt
a
t
i
on
a
nd
de
e
p
l
e
a
r
ni
ng
ba
s
e
d
e
xt
r
a
c
t
i
v
e
Q
A
(
us
i
ng
T
S
H
D
)
c
a
n he
l
p i
nf
or
m
a
t
i
on e
xt
r
a
c
t
i
on s
y
s
t
e
m
s
a
c
hi
e
v
e
t
he
s
t
a
t
e
of
t
he
a
r
t
. P
a
r
t
i
c
ul
a
r
l
y
, t
he
be
t
t
e
r
EM
r
a
t
e
(
+
7.44%
)
a
nd
F1
-
s
c
or
e
(
+
7.84%
)
i
ndi
c
a
t
e
f
oc
us
i
ng
t
he
c
ont
e
xt
i
nt
o
s
e
m
a
nt
i
c
a
l
l
y
c
ohe
r
e
nt
c
hunks
c
a
n
e
xt
r
a
c
t
m
or
e
w
i
t
h
hi
g
he
r
pr
e
c
i
s
i
on.
T
h
e
e
xpe
r
i
m
e
nt
a
l
r
e
s
ul
t
s
s
uppor
t
t
ha
t
our
T
S
H
D
-
b
a
s
e
d
s
e
gm
e
nt
a
t
i
on
r
e
m
a
r
ka
bl
y
i
m
pr
o
v
e
s
t
he
e
f
f
e
c
t
i
v
e
ne
s
s
of
t
r
a
ns
f
or
m
e
r
m
od
e
l
s
i
n
r
e
s
u
m
e
e
xt
r
a
c
t
i
on,
pr
o
v
i
ng
t
he
hy
pot
he
s
i
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.