I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
1
5
,
No
.
2
,
A
p
r
il 2
0
2
6
,
p
p
.
1
9
1
9
~
1
9
3
0
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
15
.i
2
.
p
p
1
9
1
9
-
1
9
3
0
1919
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Summ
a
riz
a
tion
of
Indo
Sum
da
tas
et
using
e
n
ha
nced
Text
Ra
nk
with
weig
hted
wo
rd embed
ding
E
v
i Y
uli
a
nti
,
P
ia
wa
i Sa
id U
m
ba
ra
D
e
p
a
r
t
me
n
t
o
f
C
o
mp
u
t
e
r
S
c
i
e
n
c
e
,
F
a
c
u
l
t
y
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
U
n
i
v
e
r
si
t
a
s I
n
d
o
n
e
si
a
,
D
e
p
o
k
,
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
Au
g
1
5
,
2
0
2
4
R
ev
is
ed
Feb
7
,
2
0
2
6
Acc
ep
ted
Ma
r
5
,
2
0
2
6
Th
is
stu
d
y
e
v
a
lu
a
tes
th
e
e
ffe
c
ti
v
e
n
e
ss
o
f
c
o
m
b
in
in
g
th
e
Tex
tRan
k
m
e
th
o
d
with
wo
r
d
e
m
b
e
d
d
i
n
g
o
n
t
h
e
I
n
d
o
n
e
sia
n
tex
t
s
u
m
m
a
riza
ti
o
n
(
In
d
o
S
u
m
)
d
a
tas
e
t.
Two
e
x
p
e
rime
n
tal
sc
e
n
a
rio
s
we
re
a
p
p
li
e
d
:
u
n
we
i
g
h
ted
a
n
d
we
ig
h
ted
.
Th
e
u
n
we
i
g
h
te
d
sc
e
n
a
rio
i
n
c
o
rp
o
ra
tes
wo
r
d
e
m
b
e
d
d
in
g
,
su
c
h
a
s
Wo
rd
2
Ve
c
,
F
a
stTex
t
,
a
n
d
I
n
d
o
n
e
sia
n
b
id
irec
ti
o
n
a
l
e
n
c
o
d
e
r
re
p
re
se
n
tatio
n
s
fro
m
tran
sfo
rm
e
rs
(In
d
o
BER
T
)
,
in
t
o
t
h
e
Tex
tRa
n
k
fra
m
e
wo
rk
.
Th
e
we
ig
h
ted
sc
e
n
a
rio
fu
rt
h
e
r
a
u
g
m
e
n
ts
th
e
term
fre
q
u
e
n
c
y
-
i
n
v
e
rse
d
o
c
u
m
e
n
t
fre
q
u
e
n
c
y
(
TF
-
IDF)
we
ig
h
ti
n
g
t
o
th
e
wo
rd
e
m
b
e
d
d
in
g
in
t
h
e
in
it
ial
sc
e
n
a
rio
.
Ou
r
re
su
lt
s
o
n
t
h
e
e
ffe
c
ti
v
e
n
e
ss
o
f
e
n
h
a
n
c
e
d
Tex
tRan
k
u
s
in
g
wo
r
d
e
m
b
e
d
d
in
g
o
n
I
n
d
o
S
u
m
d
a
ta
a
re
c
o
n
siste
n
t
wi
th
t
h
o
se
re
p
o
rted
i
n
p
re
v
io
u
s
wo
rk
o
n
Li
p
u
ta
n
6
d
a
ta.
Bo
t
h
sc
e
n
a
rio
s
c
a
n
sig
n
ifi
c
a
n
tl
y
i
m
p
ro
v
e
th
e
e
ffe
c
ti
v
e
n
e
ss
o
f
Tex
tRan
k
s
u
m
m
a
riza
ti
o
n
.
T
he
n
,
th
e
we
ig
h
ted
sc
e
n
a
rio
sh
o
we
d
p
e
rf
o
rm
a
n
c
e
imp
ro
v
e
m
e
n
t
in
m
o
st
su
m
m
a
riza
ti
o
n
sy
ste
m
s
c
o
m
p
a
re
d
to
th
e
u
n
we
ig
h
ted
sc
e
n
a
rio
,
with
a
n
a
v
e
ra
g
e
p
e
rf
o
rm
a
n
c
e
in
c
re
a
se
o
f
5
.
5
5
%
in
re
c
a
ll
-
o
r
ien
ted
u
n
d
e
rstu
d
y
f
o
r
g
ist
in
g
e
v
a
lu
a
ti
o
n
(
ROU
G
E
)
-
1
a
n
d
9
.
9
5
%
in
ROU
G
E
-
2.
Th
i
s
re
su
lt
c
o
n
f
irms
th
e
r
o
b
u
stn
e
ss
o
f
th
e
e
n
h
a
n
c
e
d
Tex
tRan
k
wit
h
we
i
g
h
t
e
d
wo
r
d
e
m
b
e
d
d
in
g
o
n
t
h
e
In
d
o
S
u
m
d
a
ta.
Las
tl
y
,
o
u
r
stu
d
y
a
lso
h
ig
h
li
g
h
ts
th
e
imp
o
rtan
c
e
o
f
u
si
n
g
d
o
m
a
i
n
-
sp
e
c
ifi
c
train
in
g
d
a
ta t
o
o
p
t
imiz
e
su
m
m
a
riza
ti
o
n
p
e
rf
o
rm
a
n
c
e
.
K
ey
w
o
r
d
s
:
Fas
tTe
x
t
I
n
d
o
B
E
R
T
I
n
d
o
Su
m
Su
m
m
ar
izatio
n
T
ex
tR
an
k
W
eig
h
ted
wo
r
d
em
b
ed
d
in
g
W
o
r
d
2
Vec
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
:
E
v
i Y
u
lian
ti
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
,
Facu
lty
o
f
C
o
m
p
u
ter
Scie
n
ce
,
Un
iv
er
s
itas
I
n
d
o
n
esia
Dep
o
k
,
W
est J
av
a,
I
n
d
o
n
esia
E
m
ail:
ev
i.y
@
cs.u
i.a
c.
id
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
i
n
ter
n
et
h
as
b
ec
o
m
e
a
n
i
n
teg
r
al
p
a
r
t
to
I
n
d
o
n
esian
p
eo
p
le
in
th
e
p
ast
d
ec
a
d
e.
Acc
o
r
d
in
g
to
th
e
Data
R
ep
o
r
tal
s
tati
s
tic
s
in
2
0
2
3
,
th
e
n
u
m
b
e
r
o
f
in
ter
n
et
u
s
e
r
s
in
I
n
d
o
n
esia
r
ea
ch
ed
2
1
2
.
9
m
illi
o
n
in
2
0
2
3
,
ac
co
u
n
tin
g
f
o
r
ap
p
r
o
x
im
ately
7
7
.
0
2
% o
f
th
e
to
tal
p
o
p
u
latio
n
[
1
]
.
T
h
e
in
ter
n
et
h
as
also
s
h
if
ted
th
e
way
p
eo
p
le
s
ee
k
in
f
o
r
m
atio
n
,
f
r
o
m
r
ea
d
in
g
p
r
in
te
d
m
ed
ia
to
r
ea
d
i
n
g
o
n
lin
e
m
ed
ia.
On
a
v
er
ag
e
,
a
p
e
r
s
o
n
s
p
en
d
s
ar
o
u
n
d
eig
h
t to
n
in
e
h
o
u
r
s
o
n
lin
e
d
aily
,
p
r
im
a
r
ily
ac
ce
s
s
in
g
s
o
cial
m
ed
ia
,
an
d
r
ea
d
in
g
n
ews
[
1
]
.
Su
r
v
ey
s
o
v
er
th
e
p
ast
th
r
ee
y
ea
r
s
h
av
e
co
n
s
is
ten
tly
s
h
o
wn
th
at
o
n
lin
e
m
ed
ia
r
em
ai
n
s
th
e
m
o
s
t
p
r
ef
er
r
e
d
s
o
u
r
ce
o
f
i
n
f
o
r
m
a
tio
n
[
2
]
.
C
u
r
r
e
n
tly
,
th
o
u
s
an
d
s
o
f
ac
tiv
e
n
ews
p
latf
o
r
m
s
p
r
o
v
id
e
v
ar
io
u
s
in
f
o
r
m
atio
n
,
r
an
g
in
g
f
r
o
m
p
o
liti
cs,
ec
o
n
o
m
ics,
en
ter
tain
m
en
t
,
an
d
tech
n
o
l
o
g
y
.
Ho
we
v
er
,
th
is
in
cr
ea
s
in
g
am
o
u
n
t o
f
in
f
o
r
m
atio
n
also
cr
ea
tes th
e
ch
allen
g
e
o
f
in
f
o
r
m
a
tio
n
o
v
er
lo
a
d
,
wh
er
e
s
o
m
e
o
n
e
o
f
ten
h
as d
if
f
icu
lty
f
in
d
in
g
r
elev
an
t
in
f
o
r
m
atio
n
am
id
s
t
th
e
lar
g
e
am
o
u
n
t
o
f
t
ex
t
d
ata
av
ailab
le.
Me
th
o
d
s
th
at
ca
n
p
r
esen
t
th
e
p
r
im
ar
y
in
f
o
r
m
atio
n
f
r
o
m
a
t
ex
t
b
y
id
en
tif
y
in
g
an
d
ex
tr
ac
t
in
g
ess
en
tial
p
h
r
ases
o
r
s
en
ten
ce
s
ar
e
n
ee
d
ed
to
ad
d
r
ess
th
is
is
s
u
e.
On
e
co
m
m
o
n
ly
u
s
ed
m
eth
o
d
is
tex
t su
m
m
ar
izatio
n
.
T
ex
t
s
u
m
m
ar
izatio
n
is
a
task
o
f
ex
t
r
ac
tin
g
m
ea
n
in
g
f
u
l
in
f
o
r
m
atio
n
f
r
o
m
o
n
e
o
r
m
u
ltip
le
d
o
cu
m
en
ts
,
en
s
u
r
in
g
t
h
at
th
e
in
f
o
r
m
atio
n
p
r
esen
ted
r
e
m
ain
s
co
m
p
lete
[
3
]
.
B
ased
o
n
th
e
o
u
tp
u
t
f
o
r
m
,
tex
t
s
u
m
m
ar
izatio
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
2
,
Ap
r
il 2
0
2
6
:
1
9
1
9
-
1
9
3
0
1920
ca
n
b
e
d
iv
id
e
d
in
to
a
b
s
tr
ac
tiv
e
an
d
ex
tr
ac
tiv
e
[
4
]
.
Ab
s
tr
ac
tiv
e
s
u
m
m
ar
izatio
n
g
e
n
er
at
es
p
ar
ap
h
r
ases
o
f
s
en
ten
ce
s
in
th
e
in
p
u
t
te
x
t,
w
h
ile
ex
tr
ac
tiv
e
s
u
m
m
ar
izatio
n
s
elec
ts
m
ea
n
in
g
f
u
l
s
en
ten
ce
s
d
ir
ec
tly
f
r
o
m
th
e
in
p
u
t
tex
t
with
o
u
t
p
ar
a
p
h
r
asin
g
.
E
x
tr
ac
tiv
e
s
u
m
m
ar
izatio
n
is
o
f
ten
p
r
ef
er
r
ed
b
ec
au
s
e
it
is
ea
s
ier
to
im
p
lem
en
t
an
d
d
o
es n
o
t f
ac
e
t
h
e
co
m
p
lex
ch
allen
g
es o
f
s
em
an
tic
r
ep
r
es
en
tatio
n
an
d
n
atu
r
al
la
n
g
u
a
g
e
g
en
er
atio
n
[
5
]
.
On
e
o
f
th
e
m
o
s
t
p
o
p
u
l
ar
a
lg
o
r
ith
m
s
in
ex
t
r
ac
t
iv
e
s
u
m
m
a
r
iz
at
io
n
is
T
ex
t
R
an
k
[
6
]
,
wh
ich
ad
ap
t
s
Pag
e
R
an
k
a
lg
o
r
ith
m
[
7
]
to
r
a
n
k
s
en
t
en
c
e
s
wi
th
i
n
a
d
o
cu
m
en
t.
I
t
h
a
s
b
ee
n
wid
ely
u
s
ed
i
n
p
r
ev
io
u
s
r
e
s
ea
r
ch
o
n
v
ar
io
u
s
ta
s
k
s
d
u
e
to
i
t
s
p
r
o
v
en
ef
f
ec
tiv
en
e
s
s
,
s
u
ch
as
o
n
n
ew
s
ar
t
icl
e
s
u
m
m
ar
i
za
tio
n
[
8
]
–
[
1
4
]
,
b
io
m
ed
ic
al
tex
t
s
u
m
m
a
r
i
za
tio
n
[
1
5
]
,
l
eg
al
d
o
cu
m
en
t
s
u
m
m
ar
iz
at
i
o
n
[
1
6
]
,
web
d
o
cu
m
en
t
s
u
m
m
ar
iz
at
io
n
[
1
7
]
,
to
p
i
c/k
ey
w
o
r
d
e
x
tr
ac
tio
n
[
1
8
]
,
s
o
f
t
wa
r
e
d
o
cu
m
en
t
at
io
n
(
r
ele
a
s
e
n
o
te
p
r
o
d
u
ct
io
n
)
[
1
9
]
,
an
d
in
f
o
r
m
a
tio
n
cr
ed
ib
i
li
ty
a
s
s
e
s
s
m
en
t
[
2
0
]
.
Ho
w
ev
er
,
T
ex
t
R
an
k
h
a
s
lim
it
at
io
n
s
in
c
ap
tu
r
in
g
t
h
e
s
e
m
an
t
ic
r
el
at
io
n
s
h
ip
s
b
et
wee
n
s
en
ten
ce
s
b
e
ca
u
s
e
i
t
p
r
im
ar
i
ly
r
el
ie
s
o
n
wo
r
d
o
r
p
h
r
a
s
e
o
v
er
l
ap
s
wi
th
o
u
t
co
n
s
id
er
in
g
th
e
d
ee
p
er
co
n
t
ex
t.
T
o
o
v
er
co
m
e
th
e
s
e
l
i
m
it
at
io
n
s
,
s
ev
e
r
a
l
s
tu
d
ie
s
h
av
e
p
r
o
p
o
s
ed
in
co
r
p
o
r
a
tin
g
w
o
r
d
em
b
ed
d
in
g
s
in
to
T
ex
t
R
an
k
[
1
1
]
,
[
1
2
]
,
[
1
4
]
,
[
1
8
]
.
I
n
g
e
n
er
al,
t
h
is
s
tu
d
y
aim
s
t
o
ex
am
in
e
th
e
ef
f
ec
tiv
e
n
ess
o
f
T
ex
tR
an
k
m
eth
o
d
th
at
is
co
m
b
in
ed
with
weig
h
ted
wo
r
d
em
b
ed
d
in
g
,
p
r
o
p
o
s
ed
in
[
1
4
]
,
to
s
u
m
m
ar
iz
e
th
e
I
n
d
o
n
esian
tex
t
s
u
m
m
a
r
izatio
n
(
I
n
d
o
Su
m
)
d
ataset
[
2
1
]
.
W
o
r
d
em
b
e
d
d
in
g
[
2
2
]
,
[
2
3
]
is
u
s
ed
to
b
etter
esti
m
ate
th
e
s
en
ten
ce
s
im
ilar
ity
with
in
T
ex
tR
an
k
.
T
h
en
,
wo
r
d
weig
h
tin
g
,
s
u
ch
a
s
ter
m
f
r
eq
u
en
c
y
-
in
v
e
r
s
e
d
o
c
u
m
en
t
f
r
e
q
u
en
c
y
(
T
F
-
I
DF)
,
is
u
s
ed
to
weig
h
th
e
wo
r
d
em
b
e
d
d
in
g
,
em
p
h
asizin
g
th
e
im
p
o
r
tan
c
e
o
f
a
wo
r
d
with
in
a
d
o
cu
m
en
t
r
elativ
e
t
o
th
e
en
tire
co
r
p
u
s
,
f
o
cu
s
in
g
o
n
m
o
r
e
r
elev
a
n
t
wo
r
d
s
,
an
d
r
ed
u
ci
n
g
th
e
in
f
lu
e
n
c
e
o
f
o
r
d
in
a
r
y
wo
r
d
s
with
litt
le
in
f
o
r
m
ativ
e
v
alu
e
[
1
3
]
,
[
1
7
]
.
T
h
e
u
s
e
o
f
I
n
d
o
Su
m
d
ata
in
th
is
s
tu
d
y
aim
s
to
ev
alu
ate
th
e
r
o
b
u
s
tn
ess
o
f
th
e
m
eth
o
d
wh
en
it
is
ap
p
lied
to
o
th
er
d
atasets
wit
h
d
if
f
er
e
n
t
ch
ar
ac
ter
is
tics
.
Her
e,
th
e
I
n
d
o
Su
m
d
ataset
h
as
a
lar
g
er
n
u
m
b
er
o
f
d
o
cu
m
e
n
ts
an
d
a
lo
n
g
er
av
er
a
g
e
d
o
c
u
m
en
t
len
g
th
th
an
L
ip
u
tan
6
d
ataset
th
at
was
u
s
ed
i
n
th
e
o
r
ig
in
al
p
ap
e
r
(
th
e
s
tatis
tical
co
m
p
ar
is
o
n
o
f
t
h
ese
two
d
atasets
ca
n
b
e
s
ee
n
in
s
ec
tio
n
3
)
.
W
ith
th
ese
two
ch
ar
ac
ter
is
tics
,
th
e
u
s
e
o
f
I
n
d
o
Su
m
in
th
is
s
tu
d
y
is
ex
p
ec
ted
to
p
r
o
v
id
e
a
m
o
r
e
co
m
p
r
eh
e
n
s
iv
e
u
n
d
er
s
tan
d
in
g
of
th
e
e
f
f
ec
tiv
en
ess
o
f
th
e
en
h
an
ce
d
T
ex
tR
an
k
m
eth
o
d
o
n
I
n
d
o
n
esian
tex
ts
.
I
n
ad
d
itio
n
,
we
also
ex
p
lo
r
e
th
e
ef
f
ec
t
o
f
u
tili
zin
g
d
if
f
er
en
t
s
o
u
r
ce
s
o
f
d
ata
t
o
tr
ain
th
e
wo
r
d
em
b
ed
d
i
n
g
m
o
d
els
th
at
en
h
an
ce
T
e
x
tR
an
k
.
W
h
ile
in
[
1
4
]
,
th
e
W
o
r
d
2
Vec
an
d
Fas
tTe
x
t
m
o
d
els
wer
e
o
n
ly
tr
ain
ed
u
s
in
g
W
ik
ip
ed
ia
d
ata
;
in
th
is
w
o
r
k
,
we
also
u
s
ed
I
n
d
o
Su
m
[
2
1
]
a
n
d
T
em
p
o
[
2
4
]
d
ata
t
o
tr
ain
t
h
o
s
e
m
o
d
els.
T
h
is
ex
p
lo
r
atio
n
in
v
esti
g
ates
to
wh
at
ex
te
n
t
th
e
s
u
m
m
ar
izatio
n
s
y
s
tem
p
er
f
o
r
m
an
ce
d
i
f
f
er
s
wh
e
n
th
e
wo
r
d
em
b
ed
d
in
g
s
a
r
e
lear
n
ed
f
r
o
m
th
e
s
am
e
d
o
m
ain
co
r
p
u
s
,
i.e
.
,
th
e
n
ews
co
r
p
u
s
.
T
o
s
u
m
u
p
,
th
e
m
ain
co
n
t
r
ib
u
tio
n
s
o
f
th
is
p
a
p
er
ar
e
tw
o
f
o
ld
.
Firstl
y
,
we
ex
am
in
e
th
e
p
er
f
o
r
m
an
ce
o
f
T
ex
tR
an
k
m
eth
o
d
co
m
b
in
ed
wi
th
wo
r
d
em
b
ed
d
in
g
to
s
u
m
m
a
r
ize
I
n
d
o
Su
m
d
ata
in
an
u
n
weig
h
te
d
an
d
weig
h
te
d
s
ce
n
ar
io
.
T
h
is
s
tu
d
y
aim
s
to
co
n
f
ir
m
th
e
r
o
b
u
s
tn
ess
o
f
th
e
m
eth
o
d
wh
en
it
is
ap
p
lied
to
o
th
er
d
atasets
with
d
if
f
er
en
t
ch
ar
ac
ter
is
tics
.
Seco
n
d
ly
,
we
in
v
esti
g
ate
th
e
ef
f
e
ct
o
f
u
s
in
g
d
if
f
er
e
n
t
d
ata
s
o
u
r
ce
s
to
tr
ain
th
e
wo
r
d
em
b
ed
d
in
g
m
o
d
els
o
n
th
e
s
u
m
m
ar
izatio
n
s
y
s
tem
p
er
f
o
r
m
an
ce
.
T
h
is
s
tu
d
y
an
aly
ze
s
th
e
ef
f
ec
t
o
f
lear
n
in
g
wo
r
d
em
b
e
d
d
in
g
s
u
s
in
g
th
e
s
am
e
d
o
m
ai
n
co
r
p
u
s
(
as
t
h
e
d
o
cu
m
en
ts
to
b
e
s
u
m
m
ar
ized
)
o
n
th
e
s
u
m
m
ar
iz
atio
n
s
y
s
tem
p
er
f
o
r
m
an
ce
.
T
h
is
s
tu
d
y
p
r
o
v
id
es
v
alu
a
b
le
i
n
s
ig
h
ts
in
to
f
u
tu
r
e
s
u
m
m
ar
iza
tio
n
r
esear
ch
.
I
t
em
p
h
asizes
th
e
ef
f
icac
y
o
f
T
ex
tR
an
k
m
eth
o
d
th
at
is
co
m
b
in
ed
with
weig
h
ted
wo
r
d
em
b
ed
d
in
g
o
n
th
e
I
n
d
o
Su
m
d
ataset,
an
d
th
e
ef
f
ec
t
o
f
u
s
in
g
th
e
s
am
e
d
o
m
ain
co
r
p
u
s
to
tr
ain
w
o
r
d
em
b
e
d
d
in
g
m
o
d
els
o
n
th
e
s
u
m
m
ar
izatio
n
p
er
f
o
r
m
a
n
ce
.
T
h
is
r
esear
ch
also
a
d
v
an
ce
s
t
h
e
d
ev
elo
p
m
en
t
o
f
m
o
r
e
e
f
f
ec
tiv
e
I
n
d
o
n
esian
n
atu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
(
NL
P)
ap
p
licatio
n
s
,
m
o
r
e
s
p
ec
if
ically
,
a
s
u
m
m
ar
izatio
n
s
y
s
tem
,
with
im
p
licatio
n
s
f
o
r
im
p
r
o
v
i
n
g
u
s
er
ef
f
ec
tiv
en
ess
an
d
ef
f
icien
c
y
in
ac
ce
s
s
in
g
im
p
o
r
tan
t in
f
o
r
m
atio
n
f
r
o
m
d
o
cu
m
en
ts
.
T
h
e
r
est
o
f
th
e
p
ap
er
is
th
e
n
o
r
g
a
n
ized
as
f
o
llo
ws.
Sectio
n
2
p
r
esen
ts
th
e
r
ev
iews
o
f
r
elate
d
liter
atu
r
e
.
Sectio
n
3
d
escr
ib
es
th
e
m
eth
o
d
s
u
s
ed
in
o
u
r
s
tu
d
y
.
Sectio
n
4
d
escr
ib
es
o
u
r
r
esu
lts
an
d
an
aly
s
is
.
Sectio
n
s
5
an
d
6
,
r
esp
ec
tiv
ely
ex
p
lain
th
e
d
is
cu
s
s
io
n
an
d
co
n
clu
s
io
n
o
f
o
u
r
wo
r
k
.
2.
L
I
T
E
R
AT
U
RE
R
E
VI
E
W
T
h
er
e
ar
e
s
o
m
e
p
r
e
v
io
u
s
s
tu
d
i
es
th
at
ar
e
clo
s
ely
r
elate
d
to
th
is
r
esear
ch
[
1
1
]
–
[
1
4
]
,
[
1
7
]
,
[
1
8
]
.
T
h
ese
s
tu
d
ies
en
h
an
ce
d
T
ex
tR
an
k
m
eth
o
d
u
s
in
g
wo
r
d
em
b
e
d
d
in
g
an
d
/o
r
T
F
-
I
DF
weig
h
tin
g
m
eth
o
d
s
.
T
h
e
p
o
s
itio
n
o
f
o
u
r
wo
r
k
to
war
d
s
th
ese
p
r
e
v
io
u
s
s
tu
d
ies
is
s
u
m
m
ar
ized
in
T
ab
le
1
.
B
a
r
m
a
n
et
a
l
.
[
1
1
]
e
m
p
l
o
y
e
d
th
e
g
l
o
b
a
l
v
e
c
t
o
r
s
(
G
l
o
V
e
)
e
m
b
e
d
d
i
n
g
m
o
d
e
l
,
t
r
a
i
n
e
d
o
n
2014
W
i
k
i
p
e
d
ia
co
r
p
u
s
,
to
s
u
m
m
ar
ize
E
n
g
lis
h
n
ews
ar
ticles
f
r
o
m
th
e
B
B
C
n
ews
s
u
m
m
ar
y
.
T
h
ey
u
s
ed
th
e
Glo
Ve
em
b
ed
d
in
g
to
p
r
o
d
u
ce
th
e
s
en
ten
ce
v
ec
to
r
s
to
b
e
in
p
u
t
in
to
T
ex
tR
an
k
s
u
m
m
ar
izer
.
I
n
co
n
tr
ast
to
B
ar
m
an
et
a
l.
[
1
1
]
,
we
u
s
e
W
o
r
d
2
Vec
,
Fas
tTe
x
t
,
a
n
d
I
n
d
o
B
E
R
T
em
b
e
d
d
in
g
s
to
b
e
c
o
m
b
in
e
d
with
T
e
x
t
R
an
k
m
eth
o
d
f
o
r
s
u
m
m
ar
izin
g
I
n
d
o
n
esian
te
x
t.
R
an
i
an
d
L
o
b
iy
al
[
1
2
]
u
s
ed
W
o
r
d
2
Vec
em
b
e
d
d
in
g
tr
ain
e
d
o
n
th
e
Go
o
g
le
New
s
co
r
p
u
s
i
n
th
eir
p
r
o
p
o
s
ed
s
em
a
n
tic
s
u
m
m
ar
izer
u
s
in
g
lin
g
u
is
tic
an
d
s
tatis
tical
f
ea
tu
r
es.
T
h
i
s
wo
r
k
is
d
if
f
e
r
en
t
f
r
o
m
th
eir
s
in
th
at
we
u
s
e
W
o
r
d
2
Vec
in
co
m
b
in
atio
n
with
T
e
x
tR
an
k
s
u
m
m
ar
izatio
n
alg
o
r
it
h
m
.
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
u
mma
r
iz
a
tio
n
o
f I
n
d
o
S
u
m
d
a
ta
s
et
u
s
in
g
en
h
a
n
ce
d
TextR
a
n
k
w
ith
w
eig
h
ted
w
o
r
d
emb
ed
d
i
n
g
(
E
vi
Yu
lia
n
ti)
1921
T
ab
le
1
.
R
esear
ch
p
o
s
itio
n
o
f
t
h
is
wo
r
k
W
o
r
k
Ta
sk
M
e
t
h
o
d
W
o
r
d
e
mb
e
d
d
i
n
g
W
e
i
g
h
t
i
n
g
st
a
t
u
s
o
f
w
o
r
d
e
mb
e
d
d
i
n
g
U
si
n
g
TF
-
I
D
F
?
D
a
t
a
C
o
r
p
u
s
t
o
t
r
a
i
n
w
o
r
d
e
m
b
e
d
d
i
n
g
mo
d
e
l
s
La
n
g
B
a
r
ma
n
e
t
a
l
.
[
1
1
]
S
u
mm
a
r
i
z
a
t
i
o
n
Te
x
t
R
a
n
k
G
l
o
V
e
U
n
w
e
i
g
h
t
e
d
No
B
B
C
N
e
w
s
W
i
k
i
p
e
d
i
a
a
n
d
G
i
g
a
w
o
r
d
5
EN
R
a
n
i
a
n
d
Lo
b
i
y
a
l
[
1
2
]
S
u
mm
a
r
i
z
a
t
i
o
n
S
e
ma
n
t
i
c
su
mm
a
r
i
z
e
r
u
s
i
n
g
l
i
n
g
u
i
s
t
i
c
a
n
d
st
a
t
i
st
i
c
a
l
f
e
a
t
u
r
e
s
W
o
r
d
2
V
e
c
W
e
i
g
h
t
e
d
Y
e
s
D
U
C
2
0
0
7
G
o
o
g
l
e
N
e
w
s
ID
Zu
o
e
t
a
l
.
[
1
8
]
To
p
i
c
e
x
t
r
a
c
t
i
o
n
Te
x
t
R
a
n
k
W
o
r
d
2
V
e
c
U
n
w
e
i
g
h
t
e
d
No
Li
t
e
r
a
t
u
r
e
a
b
s
t
r
a
c
t
s
W
i
k
i
p
e
d
i
a
EN
G
u
a
n
e
t
a
l
.
[
1
7
]
S
u
mm
a
r
i
z
a
t
i
o
n
Te
x
t
R
a
n
k
→
k
e
y
w
o
r
d
e
x
t
r
a
c
t
i
o
n
.
TF
-
I
D
F
→
sen
t
e
n
c
e
o
r
d
e
r
i
n
g
-
-
Y
e
s
W
e
b
d
o
c
u
me
n
t
s
-
EN
A
w
a
r
e
e
t
a
l
.
[
1
3
]
S
u
mm
a
r
i
z
a
t
i
o
n
Te
x
t
R
a
n
k
-
-
Y
e
s
B
B
C
N
e
w
s
-
EN
Y
u
l
i
a
n
t
i
e
t
a
l
.
[
1
4
]
S
u
mm
a
r
i
z
a
t
i
o
n
Te
x
t
R
a
n
k
W
o
r
d
2
V
e
c
,
F
a
st
T
e
x
t
,
I
n
d
o
B
ER
T
U
n
w
e
i
g
h
t
e
d
a
n
d
W
e
i
g
h
t
e
d
Y
e
s
Li
p
u
t
a
n
6
W
i
k
i
p
e
d
i
a
ID
Th
i
s
w
o
r
k
S
u
mm
a
r
i
z
a
t
i
o
n
Te
x
t
R
a
n
k
W
o
r
d
2
V
e
c
,
F
a
st
T
e
x
t
,
I
n
d
o
B
ER
T
U
n
w
e
i
g
h
t
e
d
a
n
d
W
e
i
g
h
t
e
d
Y
e
s
I
n
d
o
S
u
m
W
i
k
i
p
e
d
i
a
Te
mp
o
I
n
d
o
S
u
m
ID
Z
u
o
et
a
l.
[
1
8
]
u
s
ed
W
o
r
d
2
Vec
to
en
h
an
ce
th
e
T
ex
tR
an
k
m
eth
o
d
to
in
co
r
p
o
r
ate
th
e
s
em
an
tic
r
elatio
n
s
h
ip
b
etwe
en
k
ey
wo
r
d
s
.
T
h
e
m
ai
n
d
if
f
er
en
ce
b
et
wee
n
o
u
r
wo
r
k
an
d
Z
u
o
et
a
l
’
s
wo
r
k
is
in
th
e
d
o
wn
s
tr
ea
m
task
.
I
n
t
h
eir
wo
r
k
,
T
e
x
tR
an
k
was
ad
o
p
ted
f
o
r
th
e
to
p
ic
ex
tr
ac
tio
n
task
,
b
u
t
in
o
u
r
wo
r
k
,
it
is
ap
p
lied
f
o
r
tex
t
s
u
m
m
ar
izatio
n
.
I
n
a
d
d
itio
n
,
T
F
-
I
DF
weig
h
tin
g
is
also
e
m
p
lo
y
e
d
in
o
u
r
wo
r
k
to
weig
h
th
e
wo
r
d
em
b
ed
d
i
n
g
s
.
Gu
a
n
et
a
l.
[
1
7
]
also
u
tili
ze
d
T
e
x
tR
an
k
m
eth
o
d
f
o
r
k
e
y
wo
r
d
ex
tr
ac
tio
n
,
alth
o
u
g
h
th
e
m
ai
n
task
o
f
th
eir
s
tu
d
y
is
o
n
tex
t
s
u
m
m
ar
izatio
n
.
T
h
ey
u
s
ed
th
e
T
ex
tR
an
k
m
eth
o
d
to
e
x
tr
ac
t
im
p
o
r
tan
t
k
e
y
wo
r
d
s
f
r
o
m
d
o
cu
m
e
n
ts
,
an
d
ea
ch
s
e
n
ten
ce
in
a
d
o
cu
m
en
t
was
s
co
r
ed
u
s
in
g
th
e
to
tal
T
F
-
I
DF
s
c
o
r
es
o
f
k
ey
w
o
r
d
s
in
th
e
s
en
ten
ce
.
T
h
e
t
o
p
-
s
co
r
in
g
s
en
ten
ce
s
ar
e
th
en
tak
en
a
s
a
s
u
m
m
ar
y
.
W
h
ile
Gu
an
e
t
a
l.
[
1
7
]
u
s
ed
th
e
TF
-
I
DF m
eth
o
d
to
weig
h
th
e
k
ey
wo
r
d
s
we
u
s
e
it to
weig
h
t
h
e
wo
r
d
e
m
b
ed
d
in
g
f
o
r
tex
t su
m
m
ar
izatio
n
.
Z
awa
r
e
et
a
l.
[
1
3
]
ap
p
lied
T
F
-
I
DF
to
s
u
m
m
a
r
ize
n
ews
f
r
o
m
th
e
B
B
C
n
ews
s
u
m
m
ar
y
.
I
n
th
eir
wo
r
k
,
th
e
T
F
-
I
DF
is
u
s
ed
to
co
m
p
u
te
ea
ch
elem
en
t
o
f
th
e
s
en
ten
ce
v
ec
to
r
s
(
b
ag
-
of
-
w
o
r
d
s
v
ec
t
o
r
s
u
s
in
g
T
F
-
I
DF)
,
wh
ile
in
th
is
r
esear
ch
,
it
is
u
s
ed
to
weig
h
th
e
wo
r
d
em
b
ed
d
in
g
s
,
an
d
th
e
av
er
ag
e
weig
h
te
d
wo
r
d
em
b
ed
d
in
g
s
will
f
o
r
m
th
e
s
en
ten
ce
v
ec
t
o
r
s
.
R
ec
en
tly
,
Yu
lian
ti
et
a
l.
[
1
4
]
in
c
o
r
p
o
r
ated
W
o
r
d
2
Vec
,
Fas
tTe
x
t
,
an
d
I
n
d
o
B
E
R
T
em
b
ed
d
in
g
in
t
o
T
ex
tR
an
k
m
eth
o
d
,
to
s
u
m
m
a
r
ize
I
n
d
o
n
esian
tex
t.
W
h
ile
th
ey
u
s
ed
L
ip
u
tan
6
d
ataset
[
8
]
in
th
is
wo
r
k
,
th
eir
m
eth
o
d
a
r
e
ev
al
u
ated
u
s
in
g
I
n
d
o
Su
m
d
ataset
[
2
1
]
th
at
h
as
s
ig
n
if
ican
tly
h
ig
h
er
n
u
m
b
er
o
f
d
o
c
u
m
en
ts
,
an
d
ea
ch
co
n
ta
n
in
g
l
o
n
g
er
tex
t
(
s
ee
T
ab
le
2
f
o
r
t
h
e
co
m
p
ar
is
o
n
o
f
th
ese
two
d
atasets
)
.
W
e
f
u
r
th
er
an
aly
s
e
th
e
ef
f
ec
t
o
f
tr
ain
in
g
wo
r
d
em
b
ed
d
in
g
m
o
d
els u
s
in
g
d
if
f
e
r
en
t d
ata
s
o
u
r
ce
s
.
3.
M
E
T
H
O
D
3
.
1
.
Da
t
a
s
et
I
n
d
o
Su
m
d
ata
[
2
1
]
is
u
s
ed
in
th
is
s
tu
d
y
to
ev
alu
ate
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
en
h
an
ce
d
T
ex
tR
an
k
m
eth
o
d
with
wo
r
d
em
b
ed
d
i
n
g
.
T
h
is
is
d
if
f
er
en
t
f
r
o
m
[
1
4
]
wh
o
u
tili
ze
d
th
e
L
i
p
u
tan
6
d
at
aset
u
s
in
g
th
e
s
am
e
m
eth
o
d
as
th
is
wo
r
k
.
T
h
e
u
n
d
er
ly
in
g
m
o
tiv
atio
n
f
o
r
u
s
in
g
I
n
d
o
Su
m
d
ataset
in
th
is
s
tu
d
y
is
b
ec
au
s
e
it
is
an
o
th
er
b
en
ch
m
a
r
k
f
o
r
I
n
d
o
n
esian
tex
t
s
u
m
m
ar
izatio
n
,
b
esid
es
L
ip
u
tan
6
d
ataset
th
at
h
as
d
if
f
er
en
t
ch
ar
ac
ter
is
tics
f
r
o
m
L
ip
u
tan
6
,
s
u
ch
as
a
lar
g
e
r
n
u
m
b
e
r
o
f
d
o
cu
m
en
ts
an
d
a
lo
n
g
er
av
er
ag
e
d
o
c
u
m
en
t
le
n
g
th
th
an
L
ip
u
tan
6
.
T
a
b
le
2
co
m
p
a
r
es th
e
s
tatis
tic
s
b
etwe
en
th
e
I
n
d
o
Su
m
a
n
d
L
ip
u
tan
6
d
atasets
in
m
o
r
e
d
etail.
T
ab
le
2
.
Statis
tical
co
m
p
ar
is
o
n
b
etwe
en
L
ip
u
tan
6
a
n
d
I
n
d
o
S
u
m
d
atasets
D
a
t
a
s
e
t
N
u
mb
e
r
o
f
d
o
c
u
me
n
t
s
A
v
e
r
a
g
e
n
u
m
b
e
r
o
f
se
n
t
e
n
c
e
s
p
e
r
d
o
c
u
me
n
t
A
v
e
r
a
g
e
n
u
m
b
e
r
o
f
w
o
r
d
s
p
e
r
d
o
c
u
me
n
t
A
v
e
r
a
g
e
n
u
m
b
e
r
o
f
w
o
r
d
s
p
e
r
s
e
n
t
e
n
c
e
Li
p
u
t
a
n
6
10
,
9
7
2
1
1
.
7
1
2
1
8
.
5
5
1
8
.
6
7
I
n
d
o
S
u
m
9
3
,
9
8
7
1
8
.
3
6
3
4
6
.
8
4
1
8
.
8
9
I
n
d
o
Su
m
c
o
n
s
is
ts
o
f
9
3
,
8
7
0
n
ews
ar
ticles
f
r
o
m
a
f
o
r
m
er
o
n
lin
e
n
ews
ag
g
r
eg
ato
r
s
ite
ca
lled
Sh
o
r
tir.
co
m
.
T
h
is
d
ataset
is
p
u
b
licly
av
ailab
le
u
n
d
er
Ap
ac
h
e
licen
s
e,
Ver
s
io
n
2
.
0
at
h
ttp
s
://g
ith
u
b
.
co
m
/k
ata
-
ai/
I
n
d
o
Su
m
.
E
ac
h
ar
ticle
in
clu
d
es
attr
ib
u
tes
s
u
ch
as
titl
e,
ca
teg
o
r
y
,
ar
ticle
co
n
ten
t,
n
ews
s
o
u
r
ce
,
s
o
u
r
ce
UR
L
,
ex
tr
ac
tiv
e
s
u
m
m
ar
y
f
la
g
(
b
o
o
l
ea
n
)
,
an
d
a
b
s
tr
ac
tiv
e
s
u
m
m
ar
y
.
W
e
u
s
ed
th
e
ex
tr
ac
tiv
e
s
u
m
m
ar
y
g
r
o
u
n
d
tr
u
th
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
2
,
Ap
r
il 2
0
2
6
:
1
9
1
9
-
1
9
3
0
1922
as
a
r
ef
e
r
en
ce
s
u
m
m
ar
y
in
o
u
r
ev
alu
atio
n
.
I
n
d
o
Su
m
is
o
r
g
a
n
ized
in
to
f
iv
e
cr
o
s
s
-
v
alid
atio
n
f
o
ld
s
,
ea
ch
d
i
v
id
ed
in
to
th
r
ee
s
u
b
s
ets:
tr
ain
,
d
ev
,
an
d
test
s
ets.
E
ac
h
f
o
ld
co
n
tain
s
1
8
,
7
7
4
n
ews
ar
ticles,
with
th
e
tr
ain
,
d
ev
,
an
d
test
s
et
s
h
av
in
g
ap
p
r
o
x
im
ately
eq
u
al
p
r
o
p
o
r
tio
n
s
(
ar
o
u
n
d
0
.
7
5
:0
.
0
5
:0
.
2
)
.
I
n
t
h
is
wo
r
k
,
th
e
test
s
et
s
f
r
o
m
ea
ch
f
o
ld
a
r
e
co
m
b
in
ed
f
o
r
e
v
alu
at
io
n
in
o
u
r
ex
p
er
im
en
t.
T
ab
le
3
d
escr
ib
es
th
e
s
tatis
tics
o
f
I
n
d
o
Su
m
d
ataset
u
s
ed
in
th
is
wo
r
k
.
Fro
m
T
ab
le
3
,
it
ca
n
b
e
o
b
s
er
v
e
d
th
at
th
e
d
o
cu
m
en
ts
in
th
e
I
n
d
o
Su
m
d
a
taset
h
av
e
ar
o
u
n
d
1
8
s
en
ten
ce
s
p
er
d
o
cu
m
e
n
t.
T
h
e
av
er
ag
e
n
u
m
b
er
o
f
wo
r
d
s
p
er
d
o
cu
m
en
t
is
ap
p
r
o
x
im
a
tely
3
4
7
,
with
ea
ch
s
en
ten
ce
co
n
tain
in
g
a
b
o
u
t
1
9
wo
r
d
s
.
T
o
en
s
u
r
e
co
n
s
is
ten
cy
a
n
d
c
o
m
p
ar
a
b
ilit
y
with
th
e
g
r
o
u
n
d
tr
u
th
s
u
m
m
ar
ies,
th
e
len
g
th
o
f
o
u
r
s
u
m
m
ar
ies will f
o
llo
w
th
e
len
g
th
o
f
th
e
g
r
o
u
n
d
t
r
u
th
s
u
m
m
ar
ies.
A
f
ew
p
r
ep
r
o
ce
s
s
in
g
s
tep
s
ar
e
p
er
f
o
r
m
e
d
on
th
e
I
n
d
o
Su
m
d
ata.
T
h
ey
in
clu
d
e
ca
s
e
f
o
ld
in
g
,
r
em
o
v
in
g
to
k
en
s
co
n
tain
in
g
non
-
al
p
h
an
u
m
er
ic
ch
ar
ac
ter
s
(
ex
ce
p
t
h
y
p
h
en
)
,
an
d
r
ep
lacin
g
n
u
m
b
e
r
to
k
en
s
with
a
s
p
ec
ial
to
k
en
@
@
NUM
@
@
.
T
h
is
p
r
e
-
p
r
o
ce
s
s
in
g
is
n
ee
d
e
d
b
e
ca
u
s
e
we
u
s
e
th
e
p
r
et
r
ain
ed
W
o
r
d
2
Vec
a
n
d
/o
r
Fas
tTe
x
t
th
at
was p
r
ep
r
o
ce
s
s
ed
u
s
in
g
th
e
s
am
e
p
r
o
ce
d
u
r
e.
W
e
also
u
tili
ze
th
e
I
n
d
o
n
esia
n
W
ik
ip
ed
ia
d
atab
ase
d
u
m
p
f
r
o
m
Ma
y
2
n
d
,
2
0
2
4
(
id
wik
i
-
2
0
2
4
0
5
0
2
-
p
ag
es
-
ar
ticles.x
m
l.b
z2
)
,
to
tr
ain
W
o
r
d
2
Vec
an
d
Fas
tTe
x
t
em
b
ed
d
in
g
m
o
d
els.
T
h
e
I
n
d
o
n
esian
W
ik
ip
ed
i
a
d
ataset
was
ch
o
s
en
f
o
r
its
ex
ten
s
iv
e
s
ize
an
d
p
u
b
lic
av
ai
lab
ilit
y
,
wh
ich
alig
n
s
with
co
m
m
o
n
p
r
ac
tice
in
s
im
ilar
r
esear
ch
.
Do
c
u
m
en
t
s
c
r
ap
in
g
is
ca
r
r
ied
o
u
t
u
s
in
g
th
e
Gen
s
im
lib
r
ar
y
,
wh
ic
h
ef
f
icien
tly
p
r
o
ce
s
s
es
th
e
lar
g
e
v
o
lu
m
e
o
f
W
ik
ip
ed
ia
ar
t
icles.
T
o
ex
a
m
in
e
th
e
in
f
l
u
en
c
e
o
f
u
s
in
g
d
if
f
e
r
en
t d
ata
s
o
u
r
c
es
to
tr
ain
th
e
w
o
r
d
em
b
ed
d
in
g
m
o
d
els
o
n
s
u
m
m
ar
y
p
e
r
f
o
r
m
an
ce
,
we
also
u
s
e
th
e
I
n
d
o
Su
m
d
ata
(
in
th
e
t
r
ain
an
d
v
al
s
ets)
,
co
n
s
is
tin
g
o
f
7
5
,
0
9
6
d
o
c
u
m
en
ts
,
an
d
p
r
etr
ain
ed
W
o
r
d
2
Vec
an
d
Fas
tTe
x
t
m
o
d
els
u
s
in
g
T
em
p
o
d
ata,
co
n
s
is
tin
g
o
f
4
0
0
K
d
o
cs,
1
6
0
M
wo
r
d
to
k
e
n
s
,
an
d
6
0
0
K
wo
r
d
ty
p
es.
I
n
ad
d
itio
n
t
o
W
o
r
d
2
Vec
an
d
Fas
tTe
x
t
,
we
also
lev
er
ag
e
a
p
r
e
-
tr
ain
ed
lan
g
u
a
g
e
m
o
d
el,
B
E
R
T
,
s
p
ec
if
ically
I
n
d
o
B
E
R
T
[
2
5
]
.
I
n
d
o
B
E
R
T
is
tr
ain
ed
o
n
th
e
I
n
d
o
4
B
d
ataset,
a
lar
g
e
-
s
ca
le
I
n
d
o
n
esian
c
o
r
p
u
s
t
h
at
co
n
tain
s
3
,
5
8
1
,
3
0
1
,
4
7
6
w
o
r
d
s
.
T
h
is
ex
ten
s
iv
e
d
ataset
en
s
u
r
es
th
at
th
e
B
E
R
T
m
o
d
el
ca
p
tu
r
es
a
wid
e
r
an
g
e
o
f
lin
g
u
is
tic
p
atter
n
s
an
d
n
u
an
ce
s
p
r
esen
t in
th
e
I
n
d
o
n
esian
lan
g
u
ag
e.
T
ab
le
3
.
Statis
tics
o
f
I
n
d
o
Su
m
d
atasets
S
u
b
s
e
t
N
u
mb
e
r
o
f
d
o
c
u
me
n
t
s
A
v
e
r
a
g
e
n
u
m
b
e
r
o
f
se
n
t
e
n
c
e
s
p
e
r
d
o
c
u
me
n
t
A
v
e
r
a
g
e
n
u
m
b
e
r
o
f
w
o
r
d
s
p
e
r
d
o
c
u
me
n
t
A
v
e
r
a
g
e
n
u
m
b
e
r
o
f
w
o
r
d
s
p
e
r
s
e
n
t
e
n
c
e
Tr
a
i
n
71
,
353
18
.
36
3
4
6
.
79
18
.
88
D
e
v
e
l
o
p
m
e
n
t
3
,
7
4
3
18
.
33
3
4
8
.
24
19
.
00
Te
st
18
,
774
18
.
36
3
4
6
.
77
18
.
89
To
t
a
l
93
,
870
18
.
36
3
4
6
.
84
18
.
89
3
.
2
.
T
he
f
ra
m
ewo
r
k
o
f
s
um
m
a
riza
t
i
o
n sy
s
t
em
W
e
u
s
e
th
e
en
h
an
ce
d
T
ex
tR
an
k
m
eth
o
d
u
s
in
g
weig
h
ted
w
o
r
d
em
b
e
d
d
in
g
m
eth
o
d
[
1
4
]
,
ill
u
s
tr
ated
in
Fig
u
r
e
1
.
Fo
r
d
etails
o
f
th
e
m
eth
o
d
,
p
lease
r
e
f
er
t
o
th
e
o
r
ig
in
al
p
a
p
er
.
I
n
s
u
m
m
ar
y
,
th
is
m
eth
o
d
wo
r
k
s
as
f
o
llo
ws
:
in
itially
,
wo
r
d
em
b
e
d
d
in
g
s
th
at
ar
e
o
b
tain
e
d
f
r
o
m
W
o
r
d
2
Vec
,
Fas
tTe
x
t
,
o
r
I
n
d
o
B
E
R
T
m
o
d
els
ar
e
weig
h
ted
u
s
in
g
T
F
-
I
DF
s
co
r
es
th
at
ar
e
co
m
p
u
ted
u
s
in
g
I
n
d
o
Su
m
d
ata.
T
h
en
,
s
en
ten
ce
em
b
ed
d
in
g
s
ar
e
g
en
er
ated
b
y
av
er
a
g
in
g
t
h
e
wo
r
d
em
b
ed
d
i
n
g
s
o
f
wo
r
d
s
co
n
tain
ed
i
n
th
e
s
en
ten
ce
.
T
h
e
p
air
wis
e
co
s
in
e
s
im
ilar
ity
b
etwe
en
s
en
ten
ce
s
is
co
m
p
u
te
d
to
b
u
ild
a
s
im
ilar
ity
m
atr
ix
to
b
e
in
p
u
t
in
to
th
e
T
ex
tR
an
k
m
eth
o
d
,
wh
ich
will
r
an
k
th
e
s
en
ten
ce
s
ac
co
r
d
in
g
to
th
e
r
elatio
n
s
h
i
p
b
etwe
en
o
th
er
s
en
ten
ce
s
in
a
d
o
cu
m
e
n
t.
T
h
e
to
p
N
s
en
ten
ce
s
ar
e
th
en
s
elec
ted
as a
s
u
m
m
ar
y
.
3
.
2
.
1
.
Wo
rd
em
bedd
ing
T
h
is
s
tu
d
y
ex
p
lo
r
es
th
r
ee
ty
p
e
s
o
f
wo
r
d
e
m
b
ed
d
in
g
s
:
W
o
r
d
2
Vec
[
2
2
]
,
Fas
tTe
x
t
[
2
3
]
,
an
d
I
n
d
o
B
E
R
T
[
2
5
]
.
W
e
tr
ain
th
e
W
o
r
d
2
Ve
c
an
d
Fas
tTe
x
t
m
o
d
els
u
s
in
g
I
n
d
o
n
esian
W
ik
ip
ed
ia
,
an
d
I
n
d
o
Su
m
d
ataset.
Ad
d
itio
n
ally
,
we
also
u
s
ed
p
r
e
-
tr
ain
ed
W
o
r
d
2
Vec
an
d
Fas
t
T
ex
t
m
o
d
els
lear
n
e
d
f
r
o
m
T
e
m
p
o
[
2
4
]
d
ata,
an
d
th
e
I
n
d
o
n
esian
B
E
R
T
p
r
e
-
tr
ai
n
ed
m
o
d
el
f
r
o
m
p
r
e
v
io
u
s
wo
r
k
:
I
n
d
o
B
E
R
T
[
2
5
]
.
T
h
is
wo
r
k
em
p
lo
y
s
th
e
s
k
ip
-
g
r
am
m
o
d
el
f
o
r
W
o
r
d
2
Vec
b
ec
au
s
e
it
is
m
o
r
e
ef
f
ec
tiv
e
at
ca
p
tu
r
in
g
d
etailed
s
em
an
tic
r
elatio
n
s
h
ip
s
.
Usi
n
g
th
e
Gen
s
im
lib
r
ar
y
,
t
h
e
W
o
r
d
2
Vec
m
o
d
el
is
tr
ain
ed
o
n
th
e
I
n
d
o
n
esia
n
W
ik
ip
ed
ia
an
d
I
n
d
o
Su
m
d
atas
et.
Fo
llo
win
g
a
r
esear
ch
b
y
Y
u
lian
ti
et
a
l.
[
1
4
]
,
d
ef
au
lt
h
y
p
er
p
ar
am
eter
v
al
u
es
ar
e
u
s
ed
to
b
u
ild
th
e
W
o
r
d
2
V
ec
m
o
d
el,
s
u
ch
as:
a
lear
n
in
g
r
ate
o
f
0
.
0
2
5
,
f
iv
e
tr
ain
i
n
g
e
p
o
ch
s
,
an
d
a
win
d
o
w
s
ize
o
f
f
iv
e.
W
e
u
s
e
ex
ac
tly
th
e
s
am
e
h
y
p
er
p
ar
am
eter
s
ettin
g
s
as
[
1
4
]
to
en
s
u
r
e
f
air
c
o
m
p
ar
is
o
n
.
A
Fas
tTe
x
t
m
o
d
el
in
c
o
r
p
o
r
a
tes
s
u
b
wo
r
d
i
n
f
o
r
m
atio
n
t
o
g
en
er
ate
wo
r
d
e
m
b
ed
d
in
g
.
T
h
er
ef
o
r
e,
th
is
ap
p
r
o
ac
h
ca
n
g
en
er
ate
e
m
b
ed
d
in
g
s
f
o
r
r
a
r
e
o
r
u
n
s
ee
n
wo
r
d
s
b
ased
o
n
th
eir
s
u
b
wo
r
d
s
,
ad
d
r
ess
in
g
th
e
out
-
of
-
v
o
ca
b
u
lar
y
p
r
o
b
lem
.
T
h
e
Fas
tTe
x
t
m
o
d
el
is
im
p
lem
en
ted
u
s
in
g
Gen
s
im
with
Sk
ip
-
g
r
am
a
r
ch
itectu
r
e
an
d
d
ef
au
lt
h
y
p
e
r
p
ar
am
eter
s
ettin
g
s
.
I
t
is
tr
ain
ed
o
n
th
e
I
n
d
o
n
esian
W
ik
ip
ed
ia
an
d
I
n
d
o
Su
m
d
atasets
,
f
o
llo
win
g
a
s
im
ilar
ap
p
r
o
ac
h
t
o
th
e
W
o
r
d
2
Vec
m
o
d
el.
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
u
mma
r
iz
a
tio
n
o
f I
n
d
o
S
u
m
d
a
ta
s
et
u
s
in
g
en
h
a
n
ce
d
TextR
a
n
k
w
ith
w
eig
h
ted
w
o
r
d
emb
ed
d
i
n
g
(
E
vi
Yu
lia
n
ti)
1923
B
E
R
T
g
en
er
ates
em
b
ed
d
in
g
s
b
y
co
n
s
id
er
in
g
t
h
e
co
n
tex
t
in
wh
ich
a
wo
r
d
ap
p
ea
r
s
.
T
h
is
b
i
d
ir
ec
tio
n
al
ap
p
r
o
ac
h
en
a
b
les
B
E
R
T
to
ca
p
tu
r
e
c
o
m
p
lex
wo
r
d
r
elatio
n
s
h
ip
s
an
d
c
o
n
tex
tu
al
n
u
a
n
ce
s
m
o
r
e
ef
f
ec
tiv
ely
.
I
n
th
is
wo
r
k
,
B
E
R
T
is
o
n
ly
u
s
ed
to
g
en
er
ate
wo
r
d
em
b
e
d
d
in
g
s
,
w
h
ich
a
r
e
f
u
r
th
er
in
c
o
r
p
o
r
ate
d
in
t
o
th
e
T
ex
tR
an
k
alg
o
r
ith
m
.
B
E
R
T
h
as
b
ee
n
p
r
e
-
tr
ain
ed
o
n
a
lar
g
e
I
n
d
o
n
esian
tex
t
co
r
p
u
s
o
f
ar
o
u
n
d
4
b
illi
o
n
wo
r
d
s
,
k
n
o
wn
as
I
n
d
o
B
E
R
T
,
in
two
v
er
s
io
n
s
:
I
n
d
o
B
E
R
T
B
ASE
an
d
I
n
d
o
B
E
R
T
L
AR
GE
.
Her
e,
t
h
e
lar
g
e
v
e
r
s
io
n
h
as
a
h
ig
h
er
n
u
m
b
er
o
f
p
ar
a
m
eter
s
th
an
th
e
b
ase
v
er
s
io
n
.
3
.
2
.
2
.
T
er
m
f
re
qu
ency
-
inv
er
s
e
do
cum
ent
f
re
qu
ency
weig
hting
TF
-
I
DF
is
a
m
eth
o
d
u
s
ed
to
ass
ig
n
weig
h
t
to
wo
r
d
s
in
a
tex
t
b
ased
o
n
th
eir
f
r
eq
u
e
n
cy
with
in
a
d
o
cu
m
e
n
t
an
d
ac
r
o
s
s
a
co
llectio
n
o
f
d
o
c
u
m
en
ts
.
I
n
o
u
r
s
u
m
m
ar
izatio
n
m
eth
o
d
,
T
F
-
I
DF
is
u
s
ed
to
weig
h
th
e
wo
r
d
e
m
b
ed
d
in
g
s
(
s
ee
Fig
u
r
e
1
)
,
b
y
m
u
ltip
ly
in
g
th
e
wo
r
d
em
b
ed
d
in
g
with
th
e
T
F
-
I
DF
o
f
th
at
wo
r
d
,
b
ef
o
r
e
av
er
ag
in
g
t
o
th
e
s
en
ten
ce
em
b
ed
d
in
g
.
T
h
e
av
e
r
ag
e
o
f
all
weig
h
ted
wo
r
d
em
b
ed
d
i
n
g
s
in
a
s
en
ten
ce
is
th
en
tak
en
as
a
s
en
ten
ce
v
ec
to
r
.
I
n
v
an
illa
T
ex
tR
an
k
,
th
e
s
en
ten
ce
v
ec
to
r
is
o
b
tain
ed
d
ir
ec
tly
u
s
in
g
a
s
im
p
le
b
ag
-
of
-
wo
r
d
s
m
eth
o
d
,
wh
ic
h
th
er
ef
o
r
e
d
o
n
o
t
ca
p
tu
r
e
th
e
s
em
an
tic
in
f
o
r
m
atio
n
with
in
s
en
ten
ce
s
.
I
n
th
is
wo
r
k
,
wo
r
d
em
b
e
d
d
in
g
e
n
ab
l
es
th
e
in
co
r
p
o
r
atio
n
o
f
s
em
an
tic
in
f
o
r
m
atio
n
,
a
n
d
th
e
T
F
-
I
DF
m
eth
o
d
f
u
r
t
h
er
g
iv
es
m
o
r
e
weig
h
t
to
th
e
wo
r
d
em
b
ed
d
in
g
o
f
m
o
r
e
im
p
o
r
ta
n
t
wo
r
d
s
.
T
h
er
ef
o
r
e,
th
e
s
en
ten
ce
v
ec
to
r
will
b
e
co
n
tr
ib
u
ted
m
o
r
e
b
y
th
e
m
o
r
e
im
p
o
r
tan
t w
o
r
d
s
r
ath
er
th
an
t
h
e
less
im
p
o
r
tan
t w
o
r
d
s
.
T
h
e
f
o
r
m
u
las to
co
m
p
u
te
T
F a
n
d
I
DF sco
r
es a
r
e
g
iv
e
n
b
y
(
1
)
a
n
d
(
2
)
.
(
,
)
=
(
,
)
∑
(
̂
,
)
̂
∈
(
1
)
(
)
=
|
|
|
∈
:
∈
|
(
2
)
W
h
er
e
(
,
)
is
th
e
ac
tu
al
TF
o
f
wo
r
d
t
in
d
o
c
u
m
en
t
d
,
∑
(
̂
,
)
̂
∈
is
th
e
to
ta
l
ter
m
s
in
d
o
cu
m
en
t
d
,
(
,
)
is
th
e
n
o
r
m
alize
d
T
F
o
f
wo
r
d
t
in
d
o
cu
m
e
n
t
d
,
|
|
is
th
e
to
tal
n
u
m
b
er
o
f
d
o
cu
m
e
n
ts
in
th
e
co
llectio
n
,
|
∈
:
∈
|
is
th
e
n
u
m
b
e
r
o
f
d
o
cu
m
en
ts
c
o
n
tain
in
g
t
h
e
wo
r
d
t
,
an
d
(
)
is
t
h
e
I
DF o
f
wo
r
d
t
.
Fig
u
r
e
1
.
T
h
e
f
r
a
m
ewo
r
k
o
f
th
e
en
h
an
ce
d
T
ex
tR
an
k
s
u
m
m
a
r
izatio
n
s
y
s
tem
[
1
4
]
3
.
2
.
3
.
T
ex
t
Ra
nk
I
n
th
e
tex
t
s
u
m
m
ar
izatio
n
d
o
m
ain
,
T
e
x
tR
an
k
is
a
g
r
a
p
h
-
b
ased
r
an
k
in
g
al
g
o
r
ith
m
f
o
r
s
en
ten
ce
r
an
k
in
g
.
T
h
e
r
elatio
n
s
h
ip
s
b
et
wee
n
s
en
ten
ce
s
in
a
d
o
cu
m
en
t
ar
e
th
e
m
ain
f
ac
to
r
s
th
at
co
n
tr
ib
u
te
to
th
e
s
co
r
e
ass
ig
n
ed
b
y
T
ex
tR
an
k
to
ea
c
h
s
en
ten
ce
.
I
n
th
is
s
tu
d
y
,
th
e
tr
ad
itio
n
al
b
ag
-
of
-
wo
r
d
s
v
ec
to
r
s
th
at
r
e
p
r
esen
t
s
en
ten
ce
v
ec
to
r
s
in
th
e
o
r
ig
in
a
l
T
ex
tR
an
k
is
r
ep
lace
d
with
th
e
av
er
ag
e
o
f
T
F
-
I
DF
weig
h
ted
wo
r
d
e
m
b
ed
d
in
g
s
o
f
wo
r
d
s
in
th
e
s
en
ten
ce
s
.
He
r
e,
wo
r
d
em
b
ed
d
in
g
s
ar
e
o
b
ta
in
ed
f
r
o
m
W
o
r
d
2
Vec
,
Fas
tTe
x
t
,
an
d
I
n
d
o
B
E
R
T
m
o
d
els.
T
h
is
m
o
d
if
icatio
n
is
ex
p
ec
ted
to
en
h
an
ce
th
e
ac
cu
r
ac
y
o
f
th
e
o
r
i
g
in
al
T
ex
tR
an
k
s
u
m
m
ar
izatio
n
m
eth
o
d
b
ec
a
u
s
e
wo
r
d
em
b
ed
d
in
g
s
co
n
s
id
er
s
em
an
tic
s
im
ilar
ity
b
etwe
en
s
en
ten
ce
s
,
an
d
th
e
T
F
-
I
DF
allo
ws
m
o
r
e
im
p
o
r
tan
t w
o
r
d
s
to
co
n
tr
ib
u
te
h
ig
h
e
r
in
th
e
s
en
te
n
ce
v
e
cto
r
s
.
Af
ter
s
en
ten
ce
v
ec
t
o
r
s
ar
e
f
o
r
m
ed
,
p
air
wis
e
co
s
in
e
s
im
ilar
ity
b
etwe
en
s
en
te
n
ce
s
is
co
m
p
u
ted
t
o
co
n
s
tr
u
ct
a
s
im
ilar
ity
m
atr
i
x
.
T
h
is
m
atr
ix
f
o
r
m
s
th
e
b
asis
f
o
r
cr
ea
tin
g
a
weig
h
ted
g
r
a
p
h
o
f
s
en
ten
ce
s
,
wh
er
e
a
v
er
tex
is
r
ep
r
esen
ted
as
a
s
en
ten
ce
,
an
d
an
e
d
g
e
is
r
ep
r
ese
n
ted
as
a
s
im
ilar
ity
s
co
r
e
b
etwe
en
two
s
en
ten
ce
s
.
T
h
is
g
r
ap
h
will
b
e
p
r
o
ce
s
s
ed
b
y
th
e
T
ex
tR
an
k
m
eth
o
d
t
o
s
co
r
e
th
e
s
en
ten
ce
s
in
a
d
o
cu
m
en
t,
an
d
th
e
to
p
N
h
ig
h
est
-
s
co
r
in
g
s
en
ten
ce
s
ar
e
th
en
s
elec
ted
as
th
e
s
u
m
m
ar
y
.
Her
e,
N
is
ad
ju
s
ted
to
m
atch
th
e
ac
tu
al
len
g
th
o
f
th
e
g
r
o
u
n
d
tr
u
th
s
u
m
m
ar
y
to
en
s
u
r
e
f
air
r
ec
all
-
o
r
ie
n
ted
u
n
d
er
s
tu
d
y
f
o
r
g
is
tin
g
ev
al
u
atio
n
(
R
OUGE
)
s
co
r
e
co
m
p
ar
is
o
n
.
No
te
th
at
th
is
m
ay
n
o
t
r
ef
lect
th
e
r
ea
l
-
wo
r
ld
s
u
m
m
ar
izatio
n
s
ce
n
ar
io
th
at
u
s
u
ally
u
s
es
th
e
s
am
e
f
ix
ed
-
len
g
th
s
u
m
m
ar
y
f
o
r
ea
c
h
d
o
cu
m
en
t.
T
h
e
f
u
n
ctio
n
to
a
p
p
ly
t
h
e
T
ex
t
R
an
k
m
eth
o
d
is
p
r
esen
ted
in
(
3
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
2
,
Ap
r
il 2
0
2
6
:
1
9
1
9
-
1
9
3
0
1924
(
)
=
(
1
−
)
+
×
∑
∑
∈
(
)
∈
(
)
×
(
)
(
3
)
W
h
er
e
is
th
e
-
th
v
er
tex
,
(
)
is
th
e
s
co
r
e
f
o
r
th
e
-
th
v
er
tex
(t
h
e
s
en
ten
ce
s
co
r
e
)
,
(
)
is
th
e
s
et
o
f
v
er
tices
th
at
g
o
in
to
,
(
)
is
th
e
s
et
o
f
v
er
tices
th
at
g
o
o
u
t
f
r
o
m
,
is
d
am
p
in
g
f
ac
t
o
r
,
is
th
e
weig
h
t o
f
th
e
e
d
g
e
b
etwe
en
s
en
ten
ce
s
an
d
,
an
d
is
th
e
weig
h
t o
f
th
e
e
d
g
e
b
etwe
en
s
en
ten
ce
s
an
d
.
T
h
e
Py
th
o
n
lib
r
ar
y
Netwo
r
k
X
was
u
s
ed
to
ap
p
ly
th
e
T
ex
tR
an
k
m
et
h
o
d
.
W
e
s
et
th
e
d
am
p
in
g
f
ac
to
r
to
0
.
8
5
,
b
ec
au
s
e
it
h
as
b
ee
n
s
h
o
wn
to
p
er
f
o
r
m
well
in
th
e
o
r
ig
i
n
al
T
ex
tR
an
k
m
eth
o
d
[
9
]
,
as
well
as
s
o
m
e
p
r
ev
io
u
s
wo
r
k
o
n
s
u
m
m
ar
iz
atio
n
[
1
0
]
,
[
1
1
]
,
[
1
4
]
,
[
1
5
]
,
[
1
9
]
,
[
2
6
]
.
W
e
d
o
n
o
t
ex
p
er
im
en
t
with
v
ar
y
in
g
n
u
m
b
er
s
o
f
d
am
p
in
g
f
ac
to
r
s
a
n
d
d
ec
id
e
to
u
s
e
th
e
s
am
e
v
alu
e
as
[
1
4
]
,
i.e
.
,
0
.
8
5
,
t
o
en
s
u
r
e
f
air
co
m
p
ar
is
o
n
.
3
.
3
.
E
v
a
lua
t
i
o
n m
et
ric
T
h
e
r
esu
lts
ar
e
ev
alu
ated
u
s
in
g
th
e
R
OUGE
m
etr
ic
[
1
0
]
,
a
wid
ely
u
s
ed
m
etr
ic
f
o
r
ass
e
s
s
in
g
ex
tr
ac
tiv
e
tex
t
s
u
m
m
ar
izatio
n
p
er
f
o
r
m
a
n
ce
.
T
o
m
ea
s
u
r
e
o
u
r
R
OUGE
s
co
r
es,
we
ch
o
o
s
e
t
h
e
F1
-
s
co
r
e,
wh
ich
b
alan
ce
s
p
r
ec
is
io
n
an
d
r
ec
all
,
th
u
s
o
f
f
er
in
g
a
co
m
p
r
eh
e
n
s
iv
e
ass
e
s
s
m
en
t o
f
s
u
m
m
ar
izatio
n
q
u
ality
.
T
h
e
u
s
e
o
f
R
OUGE
-
F
-
1
allo
ws
f
o
r
th
e
b
alan
ce
b
etwe
en
in
f
o
r
m
ativ
en
e
s
s
an
d
co
n
cisen
ess
in
th
e
m
ea
s
u
r
em
en
t.
I
n
t
h
is
s
tu
d
y
,
we
o
n
ly
u
s
e
R
OUGE
s
co
r
e
an
d
d
o
n
o
t
co
n
s
id
er
o
t
h
er
s
u
m
m
ar
izatio
n
m
etr
ics,
s
u
ch
as
B
E
R
T
Sco
r
e
[
2
7
]
o
r
m
etr
ic
f
o
r
ev
al
u
atio
n
o
f
t
r
a
n
s
latio
n
with
ex
p
licit
o
r
d
e
r
in
g
(
ME
T
E
OR
)
[
2
8
]
,
b
ec
au
s
e
we
wan
t
to
m
ak
e
f
air
co
m
p
ar
is
o
n
with
th
e
r
esu
lts
o
b
tain
ed
in
[
1
4
]
th
at
u
s
ed
L
ip
u
tan
6
d
ataset.
No
te
th
at
th
e
g
o
al
o
f
th
is
s
tu
d
y
is
to
ev
alu
ate
th
e
r
o
b
u
s
tn
ess
o
f
th
e
m
eth
o
d
p
r
o
p
o
s
ed
in
[
1
4
]
wh
en
it
is
ev
alu
ate
d
on
I
n
d
o
Su
m
da
taset.
W
e
u
tili
ze
th
e
Py
th
o
n
lib
r
a
r
y
r
o
u
g
e
_
s
co
r
e
to
co
m
p
u
te
th
e
R
OUGE
s
co
r
es
o
f
o
u
r
g
en
er
ate
d
s
u
m
m
ar
ies.
B
ef
o
r
e
th
e
R
OUGE
s
co
r
es
ar
e
co
m
p
u
te
d
,
ca
s
e
n
o
r
m
aliza
tio
n
is
ap
p
lied
to
co
n
v
er
t
all
letter
s
in
th
e
g
r
o
u
n
d
tr
u
th
s
u
m
m
ar
ies an
d
s
y
s
tem
s
u
m
m
a
r
ies in
to
lo
wer
ca
s
e
.
T
h
e
s
tatis
tical
s
ig
n
if
ican
ce
test
b
etwe
en
th
e
R
OUGE
s
c
o
r
es
o
f
o
u
r
s
u
m
m
ar
ies
ag
ain
s
t
th
o
s
e
o
f
b
aselin
e
s
u
m
m
ar
ies
is
co
n
d
u
cted
u
s
in
g
p
air
e
d
t
-
test
.
T
h
e
co
n
f
id
en
ce
lev
el
u
s
ed
in
t
h
is
s
tu
d
y
is
0
.
0
5
.
T
h
e
s
am
p
le
s
ize
is
1
8
,
7
7
4
d
o
cu
m
e
n
ts
,
as
it
is
th
e
to
tal
n
u
m
b
er
o
f
d
o
cu
m
en
ts
in
th
e
test
s
p
lit
o
f
I
n
d
o
Su
m
d
ataset
th
at
ar
e
s
u
m
m
ar
ized
b
y
o
u
r
m
eth
o
d
s
.
Pair
ed
t
-
test
is
ch
o
s
en
b
ec
au
s
e
,
g
iv
en
p
air
s
o
f
s
co
r
es
,
i.e
.
,
s
co
r
es
o
f
o
u
r
m
eth
o
d
s
an
d
ea
ch
o
f
th
e
b
aselin
e
m
eth
o
d
s
,
we
wa
n
t
to
d
ec
i
d
e
if
th
e
d
if
f
er
en
ce
b
etwe
en
p
ai
r
ed
m
ea
s
u
r
em
e
n
ts
f
o
r
a
p
o
p
u
latio
n
is
ze
r
o
o
r
n
o
t.
I
f
th
e
d
if
f
er
e
n
ce
is
n
o
t
ze
r
o
,
th
en
it
m
ea
n
s
th
er
e
is
a
s
tati
s
tically
s
ig
n
if
ican
t
d
if
f
er
en
ce
b
etwe
en
th
e
r
esu
lts
o
f
o
u
r
s
u
m
m
ar
izatio
n
m
eth
o
d
an
d
th
e
b
aselin
e
m
eth
o
d
s
.
T
h
e
b
asic
ass
u
m
p
tio
n
s
f
o
r
p
ai
r
ed
t
-
test
[
2
9
]
ar
e
th
e
d
ata
is
co
n
tin
u
o
u
s
(
i.e
.
,
th
e
R
OUGE
s
co
r
es
ar
e
co
n
tin
u
o
u
s
v
ar
iab
les
as
t
h
ey
h
av
e
r
atio
s
ca
le)
,
ea
c
h
o
f
th
e
p
air
ed
m
ea
s
u
r
em
en
ts
ar
e
o
b
ta
in
ed
f
r
o
m
t
h
e
s
am
e
s
u
b
ject
(
ea
ch
p
air
o
f
R
OUGE
s
co
r
es
ar
e
co
m
p
u
ted
f
r
o
m
t
h
e
s
u
m
m
ar
ies
g
en
er
ated
f
r
o
m
th
e
s
am
e
d
o
cu
m
en
t)
.
T
h
e
d
if
f
e
r
en
ce
s
b
etwe
en
t
h
e
p
air
ed
o
b
s
er
v
atio
n
s
ar
e
ap
p
r
o
x
im
ately
n
o
r
m
ally
d
is
tr
ib
u
te
d
(
i.e
.
,
th
e
r
esu
lts
o
f
th
e
Sh
ap
ir
o
-
W
ilk
test
s
h
o
ws
th
at
th
e
d
if
f
er
en
c
e
o
f
R
OUGE
s
co
r
es
b
etwe
en
o
u
r
m
et
h
o
d
an
d
b
aselin
e
m
eth
o
d
s
ar
e
n
o
r
m
ally
d
is
tr
ib
u
ted
,
s
in
ce
th
e
p
-
v
alu
e
<
0
.
0
5
)
.
T
h
e
r
e
ar
e
n
o
o
u
tlier
s
ac
r
o
s
s
eith
er
o
f
t
h
e
g
r
o
u
p
(
i.e
.
,
th
e
r
e
is
n
o
R
OUGE
-
s
co
r
e
v
alu
es with
a
z
-
s
co
r
e
g
r
ea
ter
th
an
3
o
r
l
ess
th
an
-
3
)
.
4.
RE
SU
L
T
S AN
D
AN
AL
Y
SI
S
4
.
1
.
T
he
enha
nced
T
ex
t
Ra
n
k
us
ing
wo
rd
em
bedd
ing
T
ab
le
4
ex
p
lain
s
th
e
r
esu
lts
o
f
en
h
an
ce
d
T
ex
tR
an
k
m
eth
o
d
s
u
s
in
g
wo
r
d
em
b
ed
d
i
n
g
f
r
o
m
W
o
r
d
2
Vec
,
Fas
tTe
x
t
,
an
d
I
n
d
o
B
E
R
T
to
s
u
m
m
ar
ize
d
o
cu
m
e
n
ts
in
I
n
d
o
Su
m
d
ataset
(
test
s
u
b
s
et)
.
T
o
ex
am
in
e
wh
et
h
er
o
u
r
r
esu
lts
ar
e
co
n
s
is
ten
t
with
th
o
s
e
o
b
tain
e
d
in
p
r
e
v
io
u
s
w
o
r
k
u
s
in
g
L
ip
u
tan
6
d
ataset,
we
also
d
is
p
lay
in
th
e
b
o
tto
m
p
ar
t
o
f
T
ab
le
4
t
h
e
s
co
r
es
r
ep
o
r
ted
i
n
[
1
4
]
.
T
o
b
e
co
m
p
ar
ab
le,
we
o
n
ly
p
r
esen
t
i
n
t
h
e
tab
le
th
e
r
esu
lts
o
f
s
u
m
m
ar
izatio
n
s
y
s
tem
s
th
at
u
s
e
W
o
r
d
2
Vec
an
d
Fas
tTe
x
t
tr
ain
ed
u
s
in
g
W
ik
ip
ed
ia
.
T
h
e
r
esu
lts
o
f
th
o
s
e
u
s
in
g
W
o
r
d
2
Vec
an
d
Fas
tTe
x
t
tr
ain
ed
u
s
in
g
I
n
d
o
Su
m
an
d
T
em
p
o
d
ata
a
r
e
p
r
esen
ted
later
(
s
ee
T
ab
le
5
)
.
T
h
e
r
esu
lts
in
d
icate
th
at
all
s
y
s
tem
s
u
tili
zin
g
wo
r
d
em
b
ed
d
in
g
s
to
s
u
m
m
ar
ize
d
o
c
u
m
en
ts
in
I
n
d
o
Su
m
d
ataset
o
u
tp
er
f
o
r
m
t
h
e
o
r
i
g
in
al
T
e
x
tR
an
k
b
aselin
e
.
T
h
is
h
i
g
h
lig
h
ts
th
e
ef
f
ec
tiv
e
n
ess
o
f
in
teg
r
atin
g
wo
r
d
em
b
ed
d
i
n
g
s
in
to
th
e
T
e
x
tR
an
k
alg
o
r
ith
m
,
en
h
a
n
cin
g
its
s
u
m
m
ar
izatio
n
ca
p
a
b
ilit
ies.
T
h
is
is
co
n
s
is
ten
t
with
th
e
r
esu
lts
f
r
o
m
p
r
ev
io
u
s
wo
r
k
u
s
in
g
L
ip
u
tan
6
d
ataset
[
1
4
]
.
T
h
e
s
y
s
tem
th
at
g
ain
s
th
e
b
est
r
esu
lts
is
th
e
en
h
an
ce
d
T
ex
tR
an
k
u
s
in
g
I
n
d
o
B
E
R
T
L
AR
G
E
em
b
ed
d
in
g
,
wh
ich
is
also
in
lin
e
with
th
e
r
esu
lts
o
f
p
r
ev
io
u
s
wo
r
k
[
1
4
]
.
T
h
e
I
n
d
o
B
E
R
T
L
AR
GE
s
y
s
te
m
p
er
f
o
r
m
s
s
lig
h
tly
b
etter
th
an
th
e
I
n
d
o
B
E
R
T
B
ASE
s
y
s
tem
,
ac
h
iev
in
g
R
OUGE
-
1
an
d
R
OUGE
-
2
s
co
r
es
o
f
0
.
4
6
0
a
n
d
0
.
3
4
5
co
m
p
ar
e
d
to
0
.
4
3
5
a
n
d
0
.
3
1
3
,
r
esp
ec
tiv
ely
.
T
h
is
m
ar
g
in
al
im
p
r
o
v
e
m
en
t
a
lig
n
s
with
p
r
ev
io
u
s
f
in
d
in
g
s
,
lik
ely
d
u
e
t
o
I
n
d
o
B
E
R
T
L
AR
GE
's
g
r
ea
ter
ca
p
ac
it
y
to
ca
p
tu
r
e
co
n
te
x
tu
al
n
u
an
ce
s
g
iv
en
its
lar
g
er
n
u
m
b
er
o
f
lay
er
s
an
d
p
a
r
am
et
er
s
.
No
te
th
at
th
e
em
b
ed
d
in
g
d
im
en
s
io
n
ality
an
d
th
e
n
u
m
b
e
r
o
f
h
id
d
e
n
lay
er
s
,
as we
ll a
s
atten
tio
n
h
ea
d
s
in
I
n
d
o
B
E
R
T
L
AR
GE
,
ar
e
h
ig
h
er
th
an
I
n
d
o
B
E
R
T
B
A
SE.
C
o
n
s
e
q
u
en
tly
,
th
e
n
u
m
b
er
o
f
p
ar
am
eter
s
lear
n
ed
in
I
n
d
o
B
E
R
T
L
AR
GE
is
al
s
o
h
ig
h
e
r
th
an
I
n
d
o
B
E
R
T
B
ASE.
T
h
e
em
b
ed
d
in
g
s
ize
o
f
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
u
mma
r
iz
a
tio
n
o
f I
n
d
o
S
u
m
d
a
ta
s
et
u
s
in
g
en
h
a
n
ce
d
TextR
a
n
k
w
ith
w
eig
h
ted
w
o
r
d
emb
ed
d
i
n
g
(
E
vi
Yu
lia
n
ti)
1925
I
n
d
o
B
E
R
T
L
AR
GE
is
1
,
0
2
4
,
wh
ile
th
at
o
f
I
n
d
o
B
E
R
T
B
ASE
is
7
6
8
;
T
h
e
n
u
m
b
er
o
f
h
id
d
en
lay
er
s
in
I
n
d
o
B
E
R
T
L
AR
GE
is
2
4
,
w
h
ile
th
at
in
I
n
d
o
B
E
R
T
B
AS
E
is
1
2
;
T
h
e
n
u
m
b
e
r
o
f
a
tten
tio
n
h
ea
d
s
in
I
n
d
o
B
E
R
T
L
AR
GE
is
1
6
,
wh
il
e
th
at
in
I
n
d
o
B
E
R
T
B
A
SE
is
1
2
;
T
h
e
to
tal
p
ar
am
eter
s
o
f
I
n
d
o
B
E
R
T
L
AR
GE
is
3
3
5
.
2
M
,
wh
ile
th
at
o
f
I
n
d
o
B
E
R
T
B
A
SE
is
1
2
4
.
5
M.
Ov
er
all,
co
n
tex
t
u
alize
d
wo
r
d
em
b
ed
d
i
n
g
s
,
I
n
d
o
B
E
R
T
,
s
ig
n
if
ican
tly
b
o
o
s
t
th
e
p
e
r
f
o
r
m
a
n
ce
o
f
th
e
T
ex
tR
an
k
alg
o
r
ith
m
.
I
t
ca
n
ca
p
tu
r
e
m
o
r
e
s
em
an
tic
m
e
an
in
g
s
an
d
lin
g
u
is
tic
f
ea
tu
r
e
s
th
an
s
tatic
wo
r
d
em
b
ed
d
in
g
m
o
d
els
(
W
o
r
d
2
Ve
c
an
d
Fas
tTe
x
t)
.
T
h
is
im
p
r
o
v
em
en
t
ca
n
b
e
attr
ib
u
ted
to
th
e
s
u
p
er
io
r
s
em
an
tic
u
n
d
er
s
tan
d
i
n
g
p
r
o
v
id
e
d
b
y
th
e
s
e
em
b
ed
d
in
g
s
,
wh
ic
h
r
esu
lts
in
m
o
r
e
ac
c
u
r
ate
s
u
m
m
a
r
ies.
B
ased
o
n
r
u
n
n
in
g
tim
es
o
f
m
e
th
o
d
s
,
it
ap
p
ea
r
s
f
r
o
m
th
e
tab
l
e
th
at
T
ex
tR
an
k
+Wo
r
d
2
Vec
i
s
th
e
m
o
s
t
ef
f
icien
t
b
ec
au
s
e
it
ca
n
r
u
n
v
er
y
f
ast,
r
eq
u
ir
in
g
o
n
ly
5
m
in
u
tes
to
s
u
m
m
ar
ize
ar
o
u
n
d
1
9
K
d
o
cu
m
en
ts
.
T
h
e
T
ex
tR
an
k
+FastTe
x
t ta
k
es lo
n
g
er
tim
e
to
f
in
is
h
th
e
s
am
e
jo
b
b
ec
au
s
e
in
th
e
Fas
tTe
x
t m
o
d
e
l,
th
e
em
b
ed
d
i
n
g
o
f
ea
ch
wo
r
d
is
o
b
tain
e
d
b
y
s
u
m
m
in
g
u
p
th
e
em
b
e
d
d
in
g
s
o
f
its
s
u
b
wo
r
d
s
.
T
h
er
ef
o
r
e,
its
co
m
p
u
tatio
n
ta
k
es
lo
n
g
er
.
I
n
ad
d
itio
n
,
W
ik
ip
ed
ia
co
llectio
n
also
co
n
tain
s
a
h
u
g
e
n
u
m
b
e
r
o
f
wo
r
d
s
,
an
d
th
is
m
ak
es
th
e
v
o
ca
b
u
lar
y
s
ize
f
o
r
Fas
tTe
x
t
m
u
ch
b
ig
g
er
th
an
W
o
r
d
2
Vec
,
r
esu
ltin
g
in
m
u
ch
h
ig
h
e
r
ex
e
cu
tio
n
tim
e
f
o
r
t
h
e
T
ex
tR
an
k
+FastTe
x
t
m
o
d
el.
W
h
ile
in
th
e
u
n
weig
h
te
d
s
ce
n
ar
io
th
e
I
n
d
o
B
E
R
T
-
b
ased
m
o
d
els
ar
e
th
e
m
o
s
t
ef
f
ec
tiv
e
(
b
ased
o
n
R
OUGE
s
co
r
es),
it
tu
r
n
s
o
u
t
to
b
e
th
e
le
ast
ef
f
icien
t
as
it
tak
es
th
e
lo
n
g
est
tim
e
to
r
u
n
.
I
t
tak
es
ab
o
u
t
te
n
h
o
u
r
s
to
s
u
m
m
ar
ize
ar
o
u
n
d
1
9
K
d
o
c
u
m
en
t
s
.
T
h
is
ca
n
b
e
ex
p
lain
ed
b
ec
a
u
s
e
wh
en
g
en
er
atin
g
th
e
em
b
ed
d
in
g
o
f
wo
r
d
s
in
th
e
s
en
ten
ce
u
s
in
g
I
n
d
o
B
E
R
T
,
th
e
in
p
u
t
s
en
ten
ce
n
ee
d
s
to
b
e
p
ass
ed
f
o
r
war
d
th
r
o
u
g
h
t
h
e
n
eu
r
al
n
etwo
r
k
ar
ch
itectu
r
e
th
at
co
n
s
is
ts
o
f
s
ev
er
al
f
ee
d
-
f
o
r
wa
r
d
an
d
s
elf
att
en
tio
n
la
y
er
s
.
T
h
is
p
r
o
ce
s
s
is
clea
r
ly
ex
p
en
s
iv
e,
e
s
p
ec
ially
if
it is
co
n
d
u
cted
f
o
r
a
h
ig
h
n
u
m
b
e
r
o
f
lo
n
g
d
o
c
u
m
en
ts
.
I
t
is
in
ter
esti
n
g
t
o
n
o
te
th
at
th
e
o
r
ig
i
n
al
T
ex
tR
an
k
m
eth
o
d
t
u
r
n
s
o
u
t
to
b
e
q
u
ite
tim
e
co
n
s
u
m
in
g
.
I
t
tak
es
alm
o
s
t
ten
tim
es
lo
n
g
er
th
an
th
e
T
ex
tR
an
k
+Wo
r
d
2
V
ec
m
o
d
el.
T
h
is
ca
n
b
e
ex
p
lain
ed
as
f
o
llo
ws.
T
h
e
o
r
ig
in
al
T
e
x
tR
an
k
m
eth
o
d
u
s
es
b
ag
-
of
-
wo
r
d
s
v
ec
to
r
to
r
e
p
r
esen
t
a
s
en
ten
ce
v
ec
to
r
,
an
d
we
u
s
e
T
F
-
I
DF
r
ep
r
esen
tatio
n
to
r
e
p
r
esen
t
ea
ch
wo
r
d
elem
e
n
t
in
th
e
v
ec
to
r
.
T
h
e
v
ec
to
r
d
im
en
s
io
n
is
th
e
to
tal
u
n
iq
u
e
wo
r
d
s
in
th
e
d
o
cu
m
en
t.
T
h
er
ef
o
r
e,
th
e
d
im
e
n
s
io
n
o
f
v
ec
to
r
s
co
u
ld
b
e
m
o
r
e
th
an
a
th
o
u
s
an
d
,
d
e
p
en
d
s
o
n
th
e
len
g
t
h
o
f
th
e
d
o
cu
m
e
n
t.
T
h
e
len
g
th
ie
r
th
e
d
o
c
u
m
en
t,
t
h
e
h
ig
h
er
th
e
n
u
m
b
e
r
o
f
t
o
tal
u
n
iq
u
e
wo
r
d
s
in
th
at
d
o
c
u
m
en
t,
an
d
s
o
th
e
h
ig
h
er
th
e
d
im
en
s
io
n
o
f
s
en
ten
ce
v
e
cto
r
s
.
W
h
e
n
a
d
o
cu
m
en
t
h
as
a
h
ig
h
d
im
en
s
io
n
o
f
s
en
ten
ce
v
ec
to
r
s
,
it
tak
es
lo
n
g
er
tim
es
to
g
en
er
ate
th
e
v
ec
t
o
r
s
.
No
te
th
at
ev
er
y
tim
e
we
s
u
m
m
ar
iz
e
a
n
ew
d
o
cu
m
e
n
t,
th
e
s
en
ten
ce
v
ec
to
r
s
f
o
r
ea
c
h
d
o
c
u
m
en
t
n
ee
d
to
b
e
c
o
m
p
u
ted
f
r
o
m
s
cr
atch
.
T
h
is
is
d
if
f
er
en
t
f
r
o
m
th
e
T
ex
tR
an
k
+Wo
r
d
2
Vec
m
o
d
el
th
at
ju
s
t
n
ee
d
s
to
lo
o
k
u
p
th
e
wo
r
d
em
b
ed
d
i
n
g
f
r
o
m
th
e
p
r
e
tr
ain
ed
W
o
r
d
2
Vec
m
o
d
el,
an
d
tak
e
th
e
av
er
a
g
e
to
p
r
o
d
u
ce
s
en
ten
ce
v
ec
to
r
s
.
I
n
ad
d
itio
n
,
wh
en
th
e
d
im
en
s
io
n
o
f
s
en
ten
ce
v
ec
to
r
s
is
h
ig
h
,
it
also
ta
k
es
lo
n
g
e
r
ti
m
es
to
b
u
ild
th
e
s
en
ten
ce
s
im
ilar
ity
m
atr
ix
a
n
d
t
o
r
u
n
t
h
e
T
ex
tR
an
k
m
et
h
o
d
.
T
h
is
ca
u
s
es th
e
o
r
ig
in
al
T
ex
tR
an
k
m
eth
o
d
to
r
u
n
m
u
ch
s
lo
w
er
th
an
th
e
T
ex
tR
an
k
+Wo
r
d
2
Vec
m
eth
o
d
.
T
o
ex
am
in
e
th
e
ef
f
ec
t
o
f
u
s
in
g
a
v
ar
iatio
n
o
f
d
ata
to
tr
ain
wo
r
d
em
b
ed
d
in
g
m
o
d
els,
th
e
W
o
r
d
2
Vec
an
d
Fas
tTe
x
t
em
b
e
d
d
in
g
s
a
r
e
also
tr
ain
ed
u
s
in
g
I
n
d
o
Su
m
d
ata
(
tr
ain
s
u
b
s
et)
a
n
d
T
e
m
p
o
.
T
h
ese
em
b
ed
d
in
g
s
ar
e
th
en
u
s
ed
to
s
u
m
m
ar
ize
I
n
d
o
Su
m
d
ata
(
test
s
u
b
s
et)
.
B
ec
au
s
e
o
f
th
e
h
ig
h
co
m
p
lex
ity
a
n
d
r
eso
u
r
ce
s
n
ee
d
e
d
to
p
r
e
-
tr
ain
I
n
d
o
B
E
R
T
,
tr
ain
in
g
I
n
d
o
B
E
R
T
m
o
d
els
u
s
in
g
a
v
ar
iatio
n
o
f
d
ata
is
o
m
itted
i
n
th
is
in
v
esti
g
atio
n
.
T
h
e
r
esu
lts
ar
e
d
is
p
lay
e
d
in
T
ab
le
5
.
No
te
th
at
in
t
h
is
tab
le,
we
ca
n
n
o
t
p
u
t
t
h
e
s
co
r
es
in
[
1
4
]
s
in
ce
th
e
y
d
i
d
n
o
t u
s
e
a
d
if
f
er
en
t
d
ataset
o
th
er
th
an
W
ik
ip
ed
ia
to
t
r
ain
th
e
W
o
r
d
2
Vec
an
d
Fas
tTe
x
t m
o
d
els.
T
ab
le
4
.
T
h
e
r
esu
lts
u
s
in
g
T
ex
tR
an
k
co
m
b
in
ed
with
wo
r
d
e
m
b
ed
d
in
g
W
o
r
k
D
a
t
a
s
e
t
S
u
mm
a
r
i
z
a
t
i
o
n
s
y
st
e
m
D
a
t
a
t
o
t
r
a
i
n
w
o
r
d
e
mb
e
d
d
i
n
g
m
o
d
e
l
s
R
O
U
G
E
-
1
R
O
U
G
E
-
2
R
u
n
n
i
n
g
t
i
me
O
u
r
w
o
r
k
I
n
d
o
S
u
m
Te
x
t
R
a
n
k
-
0
.
3
8
2
0
.
2
4
7
5
9
m
2
8
s
Te
x
t
R
a
n
k
+
F
a
st
Te
x
t
W
i
k
i
p
e
d
i
a
0
.
4
1
7
*
0
.
2
8
8
*
9
m
4
5
s
Te
x
t
R
a
n
k
+
W
o
r
d
2
V
e
c
W
i
k
i
p
e
d
i
a
0
.
4
2
4
*
0
.
2
9
6
*
5
m
2
4
s
Te
x
t
R
a
n
k
+
I
n
d
o
B
E
R
T
BA
SE
I
n
d
o
4
B
(
p
r
e
-
t
r
a
i
n
e
d
)
0
.
4
3
5
*
0
.
3
1
3
*
9
h
5
1
m
3
9
s
Te
x
t
R
a
n
k
+
I
n
d
o
B
E
R
T
L
A
RG
E
I
n
d
o
4
B
(
p
r
e
-
t
r
a
i
n
e
d
)
0
.
4
6
0
*
0
.
3
4
5
*
10
h
4
m
1
7
s
Y
u
l
i
a
n
t
i
e
t
a
l
.
[
1
4
]
Li
p
u
t
a
n
6
Te
x
t
R
a
n
k
-
0
.
3
4
8
0
.
2
3
4
-
Te
x
t
R
a
n
+
F
a
s
t
Te
x
t
W
i
k
i
p
e
d
i
a
0
.
3
7
8
*
0
.
2
6
4
*
-
Te
x
t
R
a
n
k
+
W
o
r
d
2
V
e
c
W
i
k
i
p
e
d
i
a
0
.
3
8
2
*
0
.
2
6
9
*
-
Te
x
t
R
a
n
k
+
I
n
d
o
B
E
R
T
BA
S
E
I
n
d
o
4
B
(
p
r
e
-
t
r
a
i
n
e
d
)
0
.
3
9
3
*
0
.
2
8
4
*
-
Te
x
t
R
a
n
k
+
I
n
d
o
B
E
R
T
L
A
RG
E
I
n
d
o
4
B
(
p
r
e
-
t
r
a
i
n
e
d
)
0
.
3
9
4
*
0
.
2
8
5
*
-
S
y
mb
o
l
*
d
e
n
o
t
e
si
g
n
i
f
i
c
a
n
t
d
i
f
f
e
r
e
n
c
e
s a
g
a
i
n
s
t
Te
x
t
R
a
n
k
,
a
c
c
o
r
d
i
n
g
t
o
t
h
e
p
a
i
r
e
d
t
-
t
e
st
u
s
i
n
g
c
o
n
f
i
d
e
n
c
e
l
e
v
e
l
0
.
0
5
,
(
p
<
0
.
0
5
)
.
T
ab
le
5
.
T
h
e
r
esu
lts
u
s
in
g
v
ar
i
atio
n
o
f
d
ata
to
tr
ain
t
h
e
em
b
e
d
d
in
g
m
o
d
els
D
a
t
a
s
e
t
S
u
mm
a
r
i
z
a
t
i
o
n
s
y
st
e
m
D
a
t
a
t
o
t
r
a
i
n
w
o
r
d
e
m
b
e
d
d
i
n
g
mo
d
e
l
s
R
O
U
G
E
-
1
R
O
U
G
E
-
2
R
u
n
n
i
n
g
t
i
me
I
n
d
o
S
u
m
Te
x
t
R
a
n
k
+
F
a
st
Te
x
t
Te
mp
o
0
.
4
1
9
*
0
.
2
9
4
*
5
m
4
1
s
Te
x
t
R
a
n
k
+
W
o
r
d
2
V
e
c
Te
mp
o
0
.
4
3
1
*
0
.
3
1
1
*
5
m
3
7
s
Te
x
t
R
a
n
k
+
W
o
r
d
2
V
e
c
I
n
d
o
S
u
m
0
.
4
5
7
*
0
.
3
3
8
*
5
m
1
8
s
Te
x
t
R
a
n
k
+
F
a
st
Te
x
t
I
n
d
o
S
u
m
0
.
4
6
6
*
0
.
3
5
0
*
5
m
1
8
s
S
y
mb
o
l
*
d
e
n
o
t
e
si
g
n
i
f
i
c
a
n
t
d
i
f
f
e
r
e
n
c
e
s a
g
a
i
n
s
t
Te
x
t
R
a
n
k
,
a
c
c
o
r
d
i
n
g
t
o
t
h
e
p
a
i
r
e
d
t
-
t
e
st
(
p
<
0
.
0
5
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
2
,
Ap
r
il 2
0
2
6
:
1
9
1
9
-
1
9
3
0
1926
C
o
m
p
ar
in
g
th
e
r
esu
lts
f
r
o
m
T
ab
les
4
an
d
5
,
we
ca
n
s
ee
th
at
th
e
s
y
s
tem
s
with
W
o
r
d
2
Vec
an
d
Fas
tTe
x
t
m
o
d
els
tr
ain
ed
o
n
t
h
e
I
n
d
o
Su
m
d
ata
g
ain
h
ig
h
er
ac
cu
r
ac
y
th
an
t
h
o
s
e
tr
ain
ed
o
n
W
ik
ip
ed
ia
d
ata.
Mo
r
e
s
p
ec
if
ically
,
th
e
co
m
b
i
n
atio
n
o
f
T
ex
tR
an
k
with
th
e
Fas
tTe
x
t
m
o
d
el
tr
ain
ed
o
n
th
e
I
n
d
o
Su
m
d
ataset
ac
h
iev
es
th
e
h
ig
h
est
p
e
r
f
o
r
m
a
n
ce
.
T
h
is
d
em
o
n
s
tr
ates
th
at
u
s
in
g
tr
ain
in
g
d
ata
f
r
o
m
th
e
s
am
e
d
o
m
ain
as
th
e
test
d
ata
ca
n
im
p
r
o
v
e
s
y
s
tem
ac
cu
r
ac
y
.
T
h
ese
f
in
d
i
n
g
s
h
ig
h
lig
h
t
th
e
im
p
o
r
tan
ce
o
f
s
elec
tin
g
ap
p
r
o
p
r
iate
tr
ain
in
g
d
ata
an
d
m
o
d
els
to
a
ch
iev
e
o
p
tim
al
s
u
m
m
ar
izatio
n
r
esu
lts
.
T
h
e
s
y
s
tem
s
lev
er
ag
in
g
W
o
r
d
2
Vec
an
d
Fas
tTe
x
t
m
o
d
els
tr
ain
ed
o
n
T
em
p
o
d
ataset
also
ten
d
s
to
b
e
m
o
r
e
ac
c
u
r
ate
th
an
th
o
s
e
u
s
in
g
W
ik
ip
ed
ia
.
I
t
co
u
ld
b
e
b
ec
au
s
e
,
s
im
ilar
to
I
n
d
o
Su
m
,
T
em
p
o
is
also
a
co
ll
ec
tio
n
o
f
n
ews
ar
ticles.
T
h
er
e
f
o
r
e,
it
m
a
y
s
h
ar
e
a
lo
t
o
f
s
im
ilar
v
o
ca
b
u
lar
y
with
I
n
d
o
Su
m
.
T
h
is
in
cr
ea
s
ed
p
er
f
o
r
m
a
n
ce
u
n
d
e
r
s
co
r
es
th
e
im
p
o
r
tan
ce
o
f
u
s
in
g
d
o
m
ain
-
s
p
ec
if
ic
tr
ain
i
n
g
d
at
a
to
en
h
an
ce
th
e
q
u
ality
o
f
wo
r
d
em
b
ed
d
in
g
s
an
d
,
co
n
s
eq
u
en
tly
,
t
h
e
s
u
m
m
ar
izatio
n
r
esu
lts
.
T
o
b
etter
u
n
d
er
s
tan
d
wh
eth
e
r
th
e
T
em
p
o
a
n
d
I
n
d
o
Su
m
(
tr
a
in
)
d
ata
ar
e
ac
tu
ally
m
o
r
e
s
im
ilar
to
th
e
I
n
d
o
Su
m
(
test
)
c
o
m
p
a
r
ed
t
o
W
ik
ip
ed
ia
b
ec
au
s
e
t
h
ey
h
av
e
th
e
s
am
e
d
o
m
ain
d
ata,
we
c
o
m
p
u
te
ter
m
o
v
er
lap
b
etwe
en
ea
ch
o
f
th
ese
d
ata
wit
h
I
n
d
o
Su
m
(
test
)
d
ata.
T
h
is
o
v
er
lap
r
atio
is
s
im
p
ly
f
o
r
m
u
late
d
as
(
4
)
.
Ov
er
lap
R
atio
=
∣
∩
∣
∣
∣
(
4
)
W
h
er
e
is
s
et
o
f
u
n
iq
u
e
ter
m
s
in
d
ataset
A
;
is
s
et
o
f
u
n
iq
u
e
ter
m
s
in
d
ataset
B
;
an
d
∣
∩
∣
is
th
e
n
u
m
b
er
o
f
s
im
ilar
wo
r
d
s
in
d
a
tasets
A
an
d
B
.
Fro
m
T
ab
le
6
s
h
o
ws
th
at
th
e
o
v
er
lap
r
atio
b
etwe
en
W
ik
ip
e
d
ia
an
d
I
n
d
o
Su
m
(
test
)
is
o
b
v
io
u
s
ly
th
e
lo
west.
T
em
p
o
an
d
I
n
d
o
Su
m
(
tr
ain
)
d
ata
a
r
e
s
h
o
wn
to
co
n
t
ain
a
h
ig
h
e
r
n
u
m
b
er
o
f
o
v
er
l
ap
p
in
g
wo
r
d
s
with
I
n
d
o
Su
m
(
test
)
d
ata
b
ec
au
s
e
t
h
ey
h
a
v
e
th
e
s
am
e
d
o
m
ai
n
d
a
ta,
i.e
.
,
n
ews
ar
ticles.
T
h
is
ca
u
s
es
th
e
W
o
r
d
2
Vec
an
d
Fas
tTe
x
t
lear
n
ed
f
r
o
m
th
e
s
e
two
d
ataset
s
ar
e
m
o
r
e
ac
cu
r
ate
co
m
p
ar
ed
to
W
ik
ip
ed
ia
w
h
en
th
ey
ar
e
test
ed
to
th
e
I
n
d
o
Su
m
(
test
)
d
ata.
T
h
i
s
ex
p
lain
s
o
u
r
f
i
n
d
in
g
s
d
escr
ib
ed
in
T
ab
le
5
.
4
.
2
.
T
he
enha
nced
T
ex
t
Ra
n
k
us
ing
weig
hte
d wo
rd
em
bedd
ing
T
ab
le
7
p
r
esen
ts
th
e
r
esu
lts
o
f
en
h
a
n
ce
d
T
ex
tR
an
k
m
et
h
o
d
s
u
s
in
g
weig
h
ted
wo
r
d
em
b
ed
d
in
g
to
s
u
m
m
ar
ize
d
o
cu
m
en
ts
in
I
n
d
o
Su
m
d
ataset
(
test
s
u
b
s
et)
.
Ag
ain
,
we
also
d
is
p
lay
th
e
r
esu
lts
in
[
1
4
]
to
ex
am
i
n
e
wh
eth
er
o
u
r
r
esu
lts
u
s
in
g
I
n
d
o
Su
m
d
ata
a
r
e
co
n
s
is
ten
t
with
th
eir
r
esu
lts
u
s
in
g
L
ip
u
tan
6
d
ata.
I
t
ap
p
ea
r
s
f
r
o
m
th
e
tab
le
th
at
all
s
y
s
tem
s
u
tili
zin
g
weig
h
ted
w
o
r
d
e
m
b
ed
d
in
g
s
to
s
u
m
m
ar
ize
d
o
cu
m
en
ts
in
I
n
d
o
Su
m
d
ataset
o
u
tp
er
f
o
r
m
th
e
o
r
ig
in
al
T
ex
tR
an
k
b
aselin
e.
C
o
m
p
ar
ed
t
o
p
r
ev
io
u
s
r
esu
lts
with
o
u
t
T
F
-
I
DF
weig
h
tin
g
(
d
is
p
lay
ed
ea
r
lier
in
T
ab
le
4
)
,
th
er
e
is
a
s
u
b
s
tan
tial
im
p
r
o
v
e
m
en
t
in
R
OUGE
-
1
an
d
R
OUGE
-
2
s
co
r
es
f
o
r
th
e
s
y
s
tem
s
u
s
in
g
s
tatic
em
b
ed
d
in
g
(
W
o
r
d
2
Vec
a
n
d
Fas
tTe
x
t
)
.
No
tab
ly
,
th
e
c
o
m
b
in
atio
n
o
f
T
ex
tR
an
k
an
d
W
o
r
d
2
Vec
d
em
o
n
s
tr
ates
th
e
h
ig
h
est
p
er
f
o
r
m
a
n
ce
in
cr
ea
s
e
(
with
im
p
r
o
v
em
en
ts
o
f
1
3
.
2
1
%
f
o
r
R
OUGE
-
1
an
d
2
5
.
3
4
% f
o
r
R
OUGE
-
2
)
.
T
h
is
f
in
d
in
g
also
ag
r
ee
s
with
th
at
r
e
p
o
r
ted
in
[
1
4
]
u
s
in
g
L
ip
u
tan
6
d
ata.
T
ab
le
6
.
T
h
e
o
v
e
r
lap
r
atio
b
et
wee
n
tr
ain
in
g
d
ata
(
A)
an
d
test
in
g
d
ata
(
B
)
B
e
t
w
e
e
n
d
a
t
a
se
t
s
O
v
e
r
l
a
p
r
a
t
i
o
A
:
W
i
k
i
p
e
d
i
a
,
B
:
I
n
d
o
S
u
m
(
t
e
st
)
0
.
4
2
3
7
A
:
Te
m
p
o
,
B
:
I
n
d
o
S
u
m
(
t
e
s
t
)
0
.
6
0
3
4
A
:
I
n
d
o
S
u
m
(
t
r
a
i
n
)
,
B
:
I
n
d
o
S
u
m
(
t
e
st
)
0
.
7
8
2
9
N
o
t
e
:
t
r
a
i
n
i
n
g
d
a
t
a
(
A
)
i
n
d
i
c
a
t
e
s
t
h
e
d
a
t
a
u
s
e
d
t
o
t
r
a
i
n
W
o
r
d
2
V
e
c
a
n
d
F
a
st
T
e
x
t
m
o
d
e
l
s
T
ab
le
7
.
T
h
e
r
esu
lts
u
s
in
g
T
ex
tR
an
k
co
m
b
in
ed
with
weig
h
te
d
wo
r
d
e
m
b
ed
d
in
g
W
o
r
k
D
a
t
a
s
e
t
S
u
mm
a
r
i
z
a
t
i
o
n
s
y
st
e
m
Tr
a
i
n
i
n
g
d
a
t
a
R
O
U
G
E
-
1
R
O
U
G
E
-
2
R
u
n
n
i
n
g
t
i
me
Th
i
s
w
o
r
k
I
n
d
o
S
u
m
Te
x
t
R
a
n
k
-
0
.
3
8
2
0
.
2
4
7
59
m
2
8
s
Te
x
t
R
a
n
k
+
I
n
d
o
B
E
R
T
BA
S
E
+
TF
-
I
D
F
I
n
d
o
4
B
(
p
r
e
-
t
r
a
i
n
e
d
)
0
.
3
9
0
*
0
.
2
6
0
*
9
h
2
7
m
1
7
s
Te
x
t
R
a
n
k
+
I
n
d
o
B
E
R
T
L
A
RG
E
+
TF
-
I
D
F
I
n
d
o
4
B
(
p
r
e
-
t
r
a
i
n
e
d
)
0
.
4
1
8
*
0
.
2
9
5
*
9
h
3
5
m
2
3
s
Te
x
t
R
a
n
k
+
F
a
st
Te
x
t
+
TF
-
I
D
F
W
i
k
i
p
e
d
i
a
0
.
4
7
8
*
0
.
3
6
8
*
10
m
4
4
s
Te
x
t
R
a
n
k
+
W
o
r
d
2
V
e
c
+
TF
-
I
D
F
W
i
k
i
p
e
d
i
a
0
.
4
8
0
*
0
.
3
7
1
*
5
m
3
3
s
Y
u
l
i
a
n
t
i
e
t
a
l
.
[
1
4
]
Li
p
u
t
a
n
6
Te
x
t
R
a
n
k
-
0
.
3
4
8
0
.
2
3
4
-
Te
x
t
R
a
n
k
+
I
n
d
o
B
E
R
T
BA
S
E
+
TF
-
I
D
F
I
n
d
o
4
B
(
p
r
e
-
t
r
a
i
n
e
d
)
0
.
4
0
4
*
0
.
2
9
7
*
-
Te
x
t
R
a
n
k
+
F
a
st
Te
x
t
+
TF
-
I
D
F
W
i
k
i
p
e
d
i
a
0
.
4
0
4
*
0
.
2
9
8
*
-
Te
x
t
R
a
n
k
+
I
n
d
o
B
E
R
T
L
A
RG
E
+
TF
-
I
D
F
I
n
d
o
4
B
(
p
r
e
-
t
r
a
i
n
e
d
)
0
.
4
0
4
*
0
.
2
9
8
*
-
Te
x
t
R
a
n
k
+
W
o
r
d
2
V
e
c
+
TF
-
I
D
F
W
i
k
i
p
e
d
i
a
0
.
4
0
8
*
0
.
3
0
4
*
-
S
y
mb
o
l
*
d
e
n
o
t
e
si
g
n
i
f
i
c
a
n
t
d
i
f
f
e
r
e
n
c
e
s a
g
a
i
n
s
t
Te
x
t
R
a
n
k
,
a
c
c
o
r
d
i
n
g
t
o
t
h
e
p
a
i
r
e
d
t
-
t
e
st
(
p
<
0
.
0
5
)
.
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
u
mma
r
iz
a
tio
n
o
f I
n
d
o
S
u
m
d
a
ta
s
et
u
s
in
g
en
h
a
n
ce
d
TextR
a
n
k
w
ith
w
eig
h
ted
w
o
r
d
emb
ed
d
i
n
g
(
E
vi
Yu
lia
n
ti)
1927
Nex
t,
f
o
r
th
e
s
y
s
tem
s
u
s
in
g
co
n
tex
tu
al
em
b
e
d
d
in
g
(
I
n
d
o
B
E
R
T
)
,
th
e
in
cr
ea
s
ed
p
er
f
o
r
m
an
ce
o
f
weig
h
ted
wo
r
d
em
b
ed
d
in
g
ap
p
r
o
ac
h
o
v
er
u
n
weig
h
te
d
wo
r
d
em
b
ed
d
in
g
ap
p
r
o
ac
h
is
n
o
t
f
o
u
n
d
.
T
h
is
is
s
im
ilar
to
th
e
r
esu
lts
in
[
1
4
]
.
wh
ich
co
u
ld
n
o
t
f
in
d
s
tatis
tically
s
ig
n
if
ican
t
im
p
r
o
v
em
e
n
t
f
o
r
th
e
s
y
s
tem
s
u
s
in
g
I
n
d
o
B
E
R
T
in
L
ip
u
tan
6
d
ata,
alth
o
u
g
h
th
ey
s
till
s
h
o
wed
a
s
lig
h
t
in
cr
ea
s
e
in
th
eir
R
O
UGE
s
co
r
es.
Us
in
g
I
n
d
o
Su
m
d
ata,
th
e
s
y
s
tem
u
s
in
g
I
n
d
o
B
E
R
T
B
ASE
s
h
o
ws a
R
OUGE
-
1
s
co
r
e
d
ec
r
ea
s
e
o
f
0
.
0
4
5
(
1
0
.
3
4
%)
an
d
a
R
OUGE
-
2
s
co
r
e
d
ec
r
ea
s
e
o
f
0
.
0
5
3
(
1
6
.
9
3
%).
Similar
ly
,
th
e
s
y
s
tem
u
s
in
g
I
n
d
o
B
E
R
T
L
AR
GE
s
h
o
ws
a
R
OUGE
-
1
s
co
r
e
d
ec
r
ea
s
e
o
f
0
.
0
4
2
(
9
.
1
3
%)
an
d
a
R
OUGE
-
2
s
co
r
e
d
ec
r
e
ase
o
f
0
.
0
5
0
(
1
4
.
4
9
%).
T
h
is
in
d
icate
s
th
at
T
F
-
I
DF
weig
h
tin
g
is
n
o
t
ef
f
ec
tiv
e
f
o
r
co
n
tex
t
u
al
em
b
ed
d
in
g
s
,
b
ec
au
s
e
th
eir
in
h
er
en
t
co
n
tex
tu
al
u
n
d
er
s
tan
d
i
n
g
alr
ea
d
y
ca
p
tu
r
es
ess
en
tial
s
em
an
tic
n
u
an
ce
s
an
d
th
e
im
p
o
r
tan
ce
o
f
wo
r
d
s
ac
co
r
d
in
g
to
th
e
co
n
tex
t.
T
h
e
r
ef
o
r
e,
an
a
d
d
itio
n
al
weig
h
tin
g
th
at
ass
u
m
es
w
o
r
d
in
d
ep
en
d
en
ce
,
s
u
ch
as
T
F
-
I
DF,
in
s
tead
m
ak
es
th
e
s
em
an
tic
r
ep
r
esen
tatio
n
o
f
wo
r
d
s
less
ac
cu
r
ate.
T
h
is
r
e
s
u
lt
in
d
icate
s
th
at
th
e
en
h
an
c
ed
T
ex
tR
an
k
u
s
in
g
u
n
weig
h
ted
I
n
d
o
B
E
R
T
em
b
e
d
d
in
g
(
with
o
u
t
T
F
-
I
DF
weig
h
tin
g
)
is
alr
ea
d
y
n
ea
r
th
e
p
er
f
o
r
m
an
ce
ce
ilin
g
f
o
r
th
is
d
ataset.
T
ab
le
8
d
is
p
lay
s
th
e
r
esu
lts
u
s
in
g
v
ar
iatio
n
o
f
d
ata
t
o
tr
ain
wo
r
d
em
b
ed
d
i
n
g
m
o
d
e
ls
f
o
r
th
e
s
u
m
m
ar
izatio
n
s
y
s
tem
s
with
weig
h
ted
wo
r
d
em
b
e
d
d
in
g
ap
p
r
o
ac
h
.
Her
e,
th
e
W
o
r
d
2
Vec
an
d
Fas
tTe
x
t
em
b
ed
d
in
g
s
ar
e
t
r
ain
ed
u
s
in
g
I
n
d
o
Su
m
d
ata
(
tr
ai
n
s
u
b
s
et)
an
d
T
em
p
o
.
T
h
ese
em
b
ed
d
i
n
g
s
ar
e
th
e
n
u
s
ed
to
s
u
m
m
ar
ize
I
n
d
o
Su
m
d
ata
(
te
s
t
s
u
b
s
et)
.
T
h
e
v
ar
iatio
n
s
o
f
I
n
d
o
B
E
R
T
tr
ain
in
g
d
ata
a
r
e
o
m
itted
b
ec
au
s
e
its
p
r
e
-
tr
ain
in
g
p
r
o
ce
s
s
r
eq
u
ir
es
h
ig
h
co
m
p
le
x
ity
an
d
r
eso
u
r
ce
s
.
T
h
e
co
m
b
i
n
atio
n
o
f
T
ex
tR
an
k
with
th
e
Fas
tTe
x
t
an
d
W
o
r
d
2
Vec
em
b
ed
d
in
g
s
tr
ain
ed
o
n
th
e
I
n
d
o
Su
m
d
ata
s
et
co
n
s
is
ten
tly
s
h
o
ws
s
u
b
s
tan
tial
im
p
r
o
v
em
en
t
co
m
p
ar
ed
to
th
o
s
e
tr
ain
ed
o
n
W
ik
ip
ed
ia
.
T
h
is
em
p
h
asizes
t
h
e
im
p
o
r
tan
ce
o
f
u
s
in
g
d
o
m
ai
n
-
s
p
ec
if
ic
tr
ain
in
g
d
ata
in
en
h
an
cin
g
th
e
q
u
alit
y
o
f
wo
r
d
em
b
e
d
d
in
g
s
a
n
d
s
u
m
m
ar
izatio
n
r
esu
lts
.
T
h
e
u
s
e
o
f
T
em
p
o
d
ata,
h
o
wev
er
,
ap
p
e
ar
s
to
g
iv
e
in
c
o
n
s
is
ten
t
r
esu
lt,
s
in
ce
it
s
lig
h
tly
in
cr
ea
s
es
th
e
p
er
f
o
r
m
a
n
ce
o
f
s
y
s
tem
u
s
in
g
Fas
tTe
x
t
,
b
u
t it
s
lig
h
tly
d
ec
r
ea
s
es th
e
p
er
f
o
r
m
a
n
ce
o
f
s
y
s
tem
u
s
in
g
W
o
r
d
2
Vec
.
T
ab
le
8
.
T
h
e
r
esu
lts
u
s
in
g
a
v
a
r
iatio
n
o
f
d
ata
to
tr
ain
th
e
em
b
ed
d
in
g
m
o
d
els
D
a
t
a
s
e
t
S
u
mm
a
r
i
z
a
t
i
o
n
s
y
st
e
m
Tr
a
i
n
i
n
g
d
a
t
a
R
O
U
G
E
-
1
R
O
U
G
E
-
2
R
u
n
n
i
n
g
t
i
me
I
n
d
o
S
u
m
Te
x
t
R
a
n
k
+
W
o
r
d
2
V
e
c
+
TF
-
I
D
F
Te
mp
o
0
.
4
7
7
*
0
.
3
6
8
*
5
m
4
9
s
Te
x
t
R
a
n
k
+
F
a
st
Te
x
t
+
TF
-
I
D
F
Te
mp
o
0
.
4
8
0
*
0
.
3
7
2
*
5
m
3
6
s
Te
x
t
R
a
n
k
+
W
o
r
d
2
V
e
c
+
TF
-
I
D
F
I
n
d
o
S
u
m
0
.
4
9
4
*
0
.
3
8
8
*
5
m
2
5
s
Te
x
t
R
a
n
k
+
F
a
st
Te
x
t
+
TF
-
I
D
F
I
n
d
o
S
u
m
0
.
4
9
8
*
0
.
3
9
3
*
5
m
4
7
s
S
y
mb
o
l
*
d
e
n
o
t
e
si
g
n
i
f
i
c
a
n
t
d
i
f
f
e
r
e
n
c
e
s a
g
a
i
n
s
t
Te
x
t
R
a
n
k
,
a
c
c
o
r
d
i
n
g
t
o
t
h
e
p
a
i
r
e
d
t
-
t
e
st
(
p
<
0
.
0
5
)
.
5.
DIS
CU
SS
I
O
N
AND
F
UT
UR
E
WO
RK
T
o
p
r
o
v
i
d
e
a
clea
r
er
p
ictu
r
e
o
f
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
ev
alu
ated
m
eth
o
d
s
,
we
co
n
d
u
cte
d
a
co
m
p
ar
ativ
e
an
al
y
s
is
o
n
th
e
s
u
m
m
ar
ies
g
en
er
ated
b
y
v
ar
i
o
u
s
s
y
s
tem
s
ag
ain
s
t
th
e
g
r
o
u
n
d
tr
u
th
s
u
m
m
a
r
y
.
T
ab
le
9
s
h
o
ws
an
ex
a
m
p
le
o
f
g
en
er
ated
s
u
m
m
ar
ies
f
o
r
a
d
o
cu
m
en
t
with
id
3
4
in
th
e
test
s
u
b
s
et
o
f
I
n
d
o
Su
m
d
ataset
u
s
in
g
th
e
T
e
x
tR
an
k
m
eth
o
d
c
o
m
b
in
e
d
with
v
ar
io
u
s
wo
r
d
e
m
b
ed
d
in
g
m
o
d
els.
I
n
th
e
s
ce
n
ar
io
u
s
in
g
u
n
weig
h
ted
wo
r
d
em
b
e
d
d
in
g
s
,
th
e
s
u
m
m
ar
y
g
en
er
ate
d
b
y
t
h
e
T
ex
tR
an
k
+
Fas
tTe
x
t
(
W
ik
i
p
ed
ia
)
s
y
s
tem
d
o
es
n
o
t c
o
n
tain
an
y
s
en
ten
ce
s
f
r
o
m
th
e
g
r
o
u
n
d
tr
u
th
.
T
h
e
T
ex
tR
an
k
+Wo
r
d
2
Vec
(
W
ik
ip
ed
ia)
s
y
s
tem
p
er
f
o
r
m
s
s
lig
h
tly
b
etter
,
in
clu
d
in
g
o
n
e
s
en
ten
ce
f
r
o
m
th
e
g
r
o
u
n
d
tr
u
th
s
u
m
m
ar
y
.
T
h
e
T
ex
tR
an
k
+I
n
d
o
B
E
R
T
B
A
SE
an
d
T
ex
tR
an
k
+I
n
d
o
B
E
R
T
L
AR
GE
s
y
s
tem
s
also
o
n
ly
in
clu
d
e
o
n
e
s
en
ten
ce
f
r
o
m
th
e
g
r
o
u
n
d
tr
u
th
s
u
m
m
ar
y
,
b
u
t
th
e
y
ar
e
m
o
r
e
ef
f
ec
tiv
e
b
ec
a
u
s
e
co
n
tain
in
g
a
h
ig
h
e
r
n
u
m
b
e
r
o
f
ter
m
o
v
e
r
lap
s
with
th
e
g
r
o
u
n
d
tr
u
th
s
u
m
m
ar
y
,
r
esu
ltin
g
i
n
h
ig
h
er
R
OUGE
s
co
r
es.
T
h
is
s
h
o
wca
s
es
th
eir
s
u
p
er
io
r
ab
ilit
y
to
u
n
d
er
s
tan
d
th
e
co
n
tex
t
an
d
s
em
an
tics
o
f
th
e
tex
t.
W
h
en
th
e
Fas
tTe
x
t
an
d
W
o
r
d
2
Vec
ar
e
p
r
etr
ain
ed
u
s
in
g
I
n
d
o
Su
m
d
a
taset,
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
m
o
d
els
in
c
r
ea
s
ed
,
r
esu
ltin
g
in
b
o
th
t
h
e
R
OUGE
-
1
an
d
R
OUGE
-
2
s
co
r
es
ar
e
1
(
s
ee
T
ex
tR
an
k
+FastTe
x
t(
I
n
d
o
Su
m
)
a
n
d
T
ex
tR
an
k
+Wo
r
d
2
Vec
(
I
n
d
o
Su
m
)
)
.
T
h
is
r
ein
f
o
r
ce
s
o
u
r
f
in
d
i
n
g
th
at
u
s
in
g
th
e
s
am
e
d
o
m
ain
d
ataset
as
th
e
test
d
ata
ca
n
im
p
r
o
v
e
s
u
m
m
ar
y
ac
cu
r
ac
y
.
W
h
en
th
e
T
F
-
I
DF
weig
h
tin
g
is
in
co
r
p
o
r
ated
i
n
to
th
e
m
o
d
els,
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
s
y
s
tem
s
b
ased
o
n
s
tatic
wo
r
d
em
b
ed
d
in
g
s
also
in
cr
ea
s
ed
,
r
esu
ltin
g
b
o
th
o
f
th
e
R
OUGE
-
1
an
d
R
O
UGE
-
2
s
co
r
es
ar
e
1
(
s
ee
T
ex
tR
an
k
+FastTe
x
t(
W
ik
ip
ed
ia)
+T
F
-
I
DF
an
d
T
e
x
tR
an
k
+Wo
r
d
2
Vec
(
W
ik
ip
ed
i
a)
+T
F
-
I
DF)
.
T
h
is
h
ig
h
lig
h
ts
th
e
b
en
ef
its
o
f
th
e
weig
h
tin
g
m
et
h
o
d
f
o
r
s
tatic
em
b
ed
d
i
n
g
.
H
o
wev
er
,
t
h
e
r
esu
lts
o
f
en
h
an
ce
d
T
ex
tR
an
k
u
s
in
g
I
n
d
o
B
E
R
T
m
o
d
el
ar
e
n
o
t
im
p
r
o
v
e
d
b
y
th
e
weig
h
tin
g
m
eth
o
d
.
T
h
is
is
b
ec
au
s
e
th
e
b
id
ir
ec
tio
n
al
f
ea
tu
r
e
o
f
th
e
I
n
d
o
B
E
R
T
m
o
d
els
alr
ea
d
y
ca
p
tu
r
es
th
e
s
em
an
tic
m
ea
n
in
g
o
f
wo
r
d
s
.
T
h
er
ef
o
r
e
,
th
e
ad
d
itio
n
al
weig
h
tin
g
in
s
tead
m
ak
es
th
e
s
em
an
tic
r
ep
r
es
en
tatio
n
o
f
wo
r
d
s
less
ac
cu
r
ate.
T
o
o
v
er
c
o
m
e
th
is
lim
itatio
n
,
in
th
e
f
u
tu
r
e,
m
o
r
e
ad
v
an
ce
d
c
o
n
tex
t
u
al
em
b
e
d
d
in
g
s
f
r
o
m
a
m
u
ltil
in
g
u
al
lan
g
u
ag
e
m
o
d
el
th
at
in
clu
d
es B
ah
asa
I
n
d
o
n
esia in
th
e
p
r
etr
ain
i
n
g
d
ata
ca
n
b
e
i
n
v
esti
g
ated
.
Ov
er
all,
th
e
f
in
d
in
g
s
o
b
tain
e
d
in
th
is
s
tu
d
y
co
n
f
ir
m
th
e
r
esu
lts
r
ep
o
r
ted
in
[
1
4
]
th
at
co
m
b
in
in
g
T
ex
tR
an
k
an
d
wo
r
d
em
b
e
d
d
i
n
g
is
ef
f
ec
tiv
e
at
im
p
r
o
v
in
g
t
h
e
ac
cu
r
ac
y
o
f
a
s
u
m
m
ar
izati
o
n
s
y
s
tem
.
Fu
r
th
er
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
2
,
Ap
r
il 2
0
2
6
:
1
9
1
9
-
1
9
3
0
1928
ad
d
in
g
T
F
-
I
DF
weig
h
tin
g
ca
n
en
h
an
ce
th
e
s
tatic
em
b
ed
d
i
n
g
-
b
ased
s
y
s
tem
s
,
b
u
t
it
is
le
s
s
ef
f
ec
tiv
e
o
r
ev
en
d
etr
im
en
tal
f
o
r
c
o
n
tex
tu
al
em
b
ed
d
in
g
s
lik
e
I
n
d
o
B
E
R
T
.
I
n
ad
d
itio
n
,
o
u
r
r
esu
lts
also
d
is
co
v
er
n
ew
f
in
d
in
g
s
r
eg
ar
d
in
g
th
e
in
f
l
u
en
ce
o
f
u
s
in
g
th
e
s
am
e
d
ata
d
o
m
ain
f
o
r
t
r
ain
in
g
wo
r
d
em
b
e
d
d
in
g
m
o
d
els
o
n
th
e
ac
cu
r
ac
y
o
f
th
e
g
en
er
ate
d
s
u
m
m
ar
ies.
Ou
r
r
esu
lts
em
p
h
asize
th
e
im
p
o
r
tan
ce
o
f
u
s
in
g
d
o
m
ain
-
s
p
e
cif
ic
tr
ain
in
g
d
ata
in
en
h
an
cin
g
th
e
q
u
ality
o
f
w
o
r
d
em
b
ed
d
in
g
s
an
d
s
u
m
m
ar
izatio
n
r
esu
lts
.
T
h
e
p
o
s
itiv
e
R
OUGE
s
co
r
e
d
if
f
er
en
ce
s
b
etwe
en
o
u
r
m
eth
o
d
s
an
d
th
e
b
aselin
e
m
eth
o
d
s
in
d
icate
th
at
o
u
r
s
u
m
m
ar
ies
h
av
e
a
h
i
g
h
e
r
n
u
m
b
e
r
o
f
ter
m
o
v
er
lap
s
with
th
e
r
ef
er
en
ce
s
u
m
m
ar
ie
s
,
co
m
p
ar
ed
to
th
e
b
aselin
e
s
u
m
m
ar
ies.
W
h
ile
th
ese
r
esu
lts
r
ef
lect
th
e
q
u
ality
o
f
s
u
m
m
ar
ies
b
ased
o
n
o
f
f
lin
e
m
etr
ic,
it
d
o
es
n
o
t
n
ec
ess
ar
ily
r
ef
lect
th
e
u
s
er
s
’
j
u
d
g
m
e
n
t.
T
h
is
is
an
o
th
er
lim
itatio
n
o
f
th
is
wo
r
k
.
T
h
e
r
ef
o
r
e
,
f
u
r
th
e
r
u
s
er
s
tu
d
y
ca
n
also
b
e
p
er
f
o
r
m
e
d
in
th
e
f
u
tu
r
e
to
ass
ess
th
e
q
u
ality
o
f
s
u
m
m
ar
ies
d
ir
ec
tly
b
ased
o
n
u
s
er
s
’
p
er
s
p
ec
tiv
es.
T
h
is
ad
d
itio
n
al
in
v
esti
g
atio
n
o
n
h
u
m
an
e
v
alu
atio
n
o
f
s
u
m
m
ar
y
q
u
ality
ca
n
c
o
n
f
ir
m
th
e
r
esu
lts
o
b
tain
ed
in
th
is
wo
r
k
b
ased
o
n
au
to
m
atic
ev
alu
atio
n
.
T
ab
le
9
.
E
x
am
p
le
o
f
g
en
e
r
ate
d
s
u
m
m
ar
ies
S
y
st
e
m
S
u
mm
a
r
i
e
s
G
r
o
u
n
d
t
r
u
t
h
T
U
RK
I
–
L
e
b
i
h
d
a
ri
4
0
o
ra
n
g
,
se
b
a
g
i
a
n
b
e
s
a
r
m
a
n
t
a
n
a
n
g
g
o
t
a
u
n
i
t
p
a
su
k
a
n
k
h
u
s
u
s
T
u
r
k
i
,
d
i
j
a
t
u
h
i
h
u
k
u
m
a
n
se
u
m
u
r
h
i
d
u
p
se
t
e
l
a
h
d
i
n
y
a
t
a
k
a
n
t
e
r
b
u
k
t
i
b
e
r
sa
l
a
h
m
e
n
c
o
b
a
m
e
m
b
u
n
u
h
Pre
s
i
d
e
n
R
e
c
e
p
T
a
y
y
i
p
Erd
o
g
a
n
.
D
a
l
a
m
p
e
rs
i
d
a
n
g
a
n
d
i
M
u
g
l
a
,
k
o
t
a
d
i
T
u
r
k
i
b
a
r
a
t
d
a
y
a
h
a
ri
Ra
b
u
(
0
4
/
1
0
)
,
h
a
k
i
m
m
e
n
y
a
t
a
k
a
n
t
e
r
d
a
k
w
a
m
e
n
y
e
ra
n
g
h
o
t
e
l
d
i
r
e
so
r
w
i
s
a
t
a
t
e
p
i
l
a
u
t
d
i
M
a
rm
a
r
i
s,
y
a
n
g
m
e
n
j
a
d
i
t
e
m
p
a
t
p
e
n
g
i
n
a
p
a
n
Pre
s
i
d
e
n
Er
d
o
g
a
n
k
e
t
i
k
a
b
e
r
l
a
n
g
su
n
g
u
p
a
y
a
k
u
d
e
t
a
y
a
n
g
g
a
g
a
l
p
a
d
a
1
5
J
u
l
i
2
0
1
6
.
(
En
g
l
i
s
h
t
r
a
n
sl
a
t
i
o
n
:
TU
R
K
EY
–
M
o
r
e
t
h
a
n
4
0
i
n
d
i
v
i
d
u
a
l
s
,
m
o
st
o
f
t
h
e
m
f
o
r
mer
m
e
m
b
e
r
s
o
f
Tu
r
k
i
s
h
sp
e
c
i
a
l
f
o
r
c
e
s
u
n
i
t
s,
h
a
v
e
b
e
e
n
s
e
n
t
e
n
c
e
d
t
o
l
i
f
e
i
m
p
r
i
s
o
n
me
n
t
a
f
t
e
r
b
e
i
n
g
f
o
u
n
d
g
u
i
l
t
y
o
f
a
t
t
e
m
p
t
i
n
g
t
o
a
ss
a
ssi
n
a
t
e
P
r
e
s
i
d
e
n
t
R
e
c
e
p
Ta
y
y
i
p
Er
d
o
g
a
n
.
D
u
r
i
n
g
t
h
e
t
r
i
a
l
i
n
M
u
g
l
a
,
a
c
i
t
y
i
n
so
u
t
h
w
e
st
e
r
n
T
u
r
k
e
y
,
o
n
W
e
d
n
e
s
d
a
y
(
1
0
/
0
4
)
,
t
h
e
j
u
d
g
e
s
t
a
t
e
d
t
h
a
t
t
h
e
d
e
f
e
n
d
a
n
t
s
a
t
t
a
c
k
e
d
a
h
o
t
e
l
i
n
t
h
e
se
a
si
d
e
r
e
s
o
r
t
o
f
M
a
r
mari
s,
w
h
e
r
e
P
r
e
s
i
d
e
n
t
Er
d
o
g
a
n
w
a
s
st
a
y
i
n
g
d
u
r
i
n
g
t
h
e
f
a
i
l
e
d
c
o
u
p
a
t
t
e
m
p
t
o
n
J
u
l
y
1
5
,
2
0
1
6
.
)
Te
x
t
R
a
n
k
+
W
o
r
d
2
V
e
c
(
W
i
k
i
p
e
d
i
a
)
(
En
g
l
i
s
h
t
r
a
n
sl
a
t
i
o
n
:
TU
R
K
EY
–
M
o
r
e
t
h
a
n
4
0
i
n
d
i
v
i
d
u
a
l
s
,
m
o
st
o
f
t
h
e
m
f
o
r
mer
m
e
m
b
e
r
s
o
f
Tu
r
k
i
s
h
sp
e
c
i
a
l
f
o
r
c
e
s
u
n
i
t
s,
h
a
v
e
b
e
e
n
s
e
n
t
e
n
c
e
d
t
o
l
i
f
e
i
m
p
r
i
s
o
n
me
n
t
a
f
t
e
r
b
e
i
n
g
f
o
u
n
d
g
u
i
l
t
y
o
f
a
t
t
e
m
p
t
i
n
g
t
o
a
ss
a
ssi
n
a
t
e
P
r
e
si
d
e
n
t
R
e
c
e
p
T
a
y
y
i
p
Er
d
o
g
a
n
.
A
c
c
o
r
d
i
n
g
t
o
t
h
e
p
a
r
l
i
a
me
n
t
a
r
y
c
o
mm
i
t
t
e
e
i
n
v
e
s
t
i
g
a
t
i
n
g
t
h
e
i
n
c
i
d
e
n
t
,
3
4
sp
e
c
i
a
l
f
o
r
c
e
s
m
e
m
b
e
r
s
a
r
r
i
v
e
d
a
t
t
h
e
h
o
t
e
l
a
t
3
:
2
0
a
.
m.,
w
h
i
l
e
Er
d
o
g
a
n
l
a
n
d
e
d
i
n
I
st
a
n
b
u
l
a
t
3
:
4
0
a
.
m.)
(
R
O
U
G
E
-
1
=
0
.
5
5
7
,
R
O
U
G
E
-
2
=
0
.
4
3
3
)
Te
x
t
R
a
n
k
+
F
a
st
Te
x
t
(
W
i
k
i
p
e
d
i
a
)
(
En
g
l
i
s
h
t
r
a
n
sl
a
t
i
o
n
:
I
n
a
n
i
n
t
e
r
v
i
e
w
w
i
t
h
C
N
N
,
P
r
e
s
i
d
e
n
t
Er
d
o
g
a
n
s
a
i
d
,
“
I
f
I
h
a
d
s
t
a
y
e
d
1
0
o
r
1
5
mi
n
u
t
e
s
l
o
n
g
e
r
(
i
n
t
h
e
h
o
t
e
l
)
,
I
w
o
u
l
d
h
a
v
e
b
e
e
n
k
i
l
l
e
d
o
r
t
a
k
e
n
h
o
s
t
a
g
e
.
”
A
c
c
o
r
d
i
n
g
t
o
t
h
e
p
a
r
l
i
a
m
e
n
t
a
r
y
c
o
mm
i
t
t
e
e
i
n
v
e
s
t
i
g
a
t
i
n
g
t
h
e
i
n
c
i
d
e
n
t
,
3
4
s
p
e
c
i
a
l
f
o
r
c
e
s
m
e
m
b
e
r
s
a
r
r
i
v
e
d
a
t
t
h
e
h
o
t
e
l
a
t
3
:
2
0
a
.
m.
,
w
h
i
l
e
Er
d
o
g
a
n
l
a
n
d
e
d
i
n
I
st
a
n
b
u
l
a
t
3
:
4
0
a
.
m
.
)
(
R
O
U
G
E
-
1
=
0
.
3
0
3
,
R
O
U
G
E
-
2
=
0
.
0
3
4
)
Te
x
t
R
a
n
k
+
I
n
d
o
B
E
R
T
BA
S
E
(
En
g
l
i
s
h
t
r
a
n
sl
a
t
i
o
n
:
TU
R
K
EY
–
M
o
r
e
t
h
a
n
4
0
i
n
d
i
v
i
d
u
a
l
s
,
m
o
st
o
f
t
h
e
m
f
o
r
mer
m
e
m
b
e
r
s
o
f
Tu
r
k
i
s
h
sp
e
c
i
a
l
f
o
r
c
e
s
u
n
i
t
s,
h
a
v
e
b
e
e
n
s
e
n
t
e
n
c
e
d
t
o
l
i
f
e
i
m
p
r
i
s
o
n
me
n
t
a
f
t
e
r
b
e
i
n
g
f
o
u
n
d
g
u
i
l
t
y
o
f
a
t
t
e
m
p
t
i
n
g
t
o
a
s
sassi
n
a
t
e
P
r
e
si
d
e
n
t
R
e
c
e
p
Ta
y
y
i
p
Er
d
o
g
a
n
.
P
r
e
s
i
d
e
n
t
E
r
d
o
g
a
n
h
i
ms
e
l
f
l
e
f
t
h
i
s
r
e
so
r
t
i
n
M
a
r
mar
i
s
b
e
f
o
r
e
t
h
e
a
t
t
a
c
k
,
w
h
i
c
h
k
i
l
l
e
d
t
w
o
p
o
l
i
c
e
o
f
f
i
c
e
r
s,
w
a
s
l
a
u
n
c
h
e
d
.
)
(
R
O
U
G
E
-
1
=
0
.
5
9
5
,
R
O
U
G
E
-
2
=
0
.
5
1
4
)
Te
x
t
R
a
n
k
+
I
n
d
o
B
E
R
T
L
A
RG
E
(
En
g
l
i
s
h
t
r
a
n
sl
a
t
i
o
n
:
D
u
r
i
n
g
t
h
e
t
r
i
a
l
i
n
M
u
g
l
a
,
a
c
i
t
y
i
n
s
o
u
t
h
w
e
s
t
e
r
n
Tu
r
k
e
y
,
o
n
W
e
d
n
e
sd
a
y
(
1
0
/
0
4
)
,
t
h
e
j
u
d
g
e
s
t
a
t
e
d
t
h
a
t
t
h
e
d
e
f
e
n
d
a
n
t
s
a
t
t
a
c
k
e
d
a
h
o
t
e
l
i
n
t
h
e
sea
s
i
d
e
r
e
s
o
r
t
o
f
M
a
r
maris
,
w
h
e
r
e
P
r
e
s
i
d
e
n
t
Er
d
o
g
a
n
w
a
s
s
t
a
y
i
n
g
d
u
r
i
n
g
t
h
e
f
a
i
l
e
d
c
o
u
p
a
t
t
e
m
p
t
o
n
Ju
l
y
1
5
,
2
0
1
6
.
S
o
me
a
n
a
l
y
st
s
d
i
sm
i
sse
d
P
r
e
s
i
d
e
n
t
Er
d
o
g
a
n
's
c
l
a
i
m
t
h
a
t
h
e
h
a
d
a
l
r
e
a
d
y
l
e
f
t
t
h
e
h
o
t
e
l
b
e
f
o
r
e
t
h
e
f
i
r
i
n
g
sq
u
a
d
a
r
r
i
v
e
d
a
t
t
h
e
h
o
t
e
l
w
h
e
r
e
Er
d
o
g
a
n
a
n
d
h
i
s
f
a
mi
l
y
w
e
r
e
st
a
y
i
n
g
.
)
(
R
O
U
G
E
-
1
=
0
.
6
6
2
,
R
O
U
G
E
-
2
=
0
.
6
2
5
)
Te
x
t
R
a
n
k
+
W
o
r
d
2
V
e
c
(
I
n
d
o
S
u
m
)
S
a
me
a
s t
h
e
g
r
o
u
n
d
t
r
u
t
h
su
mm
a
r
y
(
R
O
U
G
E
-
1
=
1
,
R
O
U
G
E
-
2
=
1
)
Te
x
t
R
a
n
k
+
F
a
st
Te
x
t
(
I
n
d
o
S
u
m
)
S
a
me
a
s t
h
e
g
r
o
u
n
d
t
r
u
t
h
su
mm
a
r
y
(
R
O
U
G
E
-
1
=
1
,
R
O
U
G
E
-
2
=
1
)
Te
x
t
R
a
n
k
+
W
o
r
d
2
V
e
c
(
W
i
k
i
p
e
d
i
a
)
+
TF
-
I
D
F
S
a
me
a
s t
h
e
g
r
o
u
n
d
t
r
u
t
h
su
mm
a
r
y
(
R
O
U
G
E
-
1
=
1
,
R
O
U
G
E
-
2
=
1
)
Te
x
t
R
a
n
k
+
F
a
st
Te
x
t
(
W
i
k
i
p
e
d
i
a
)
+
TF
-
I
D
F
S
a
me
a
s t
h
e
g
r
o
u
n
d
t
r
u
t
h
su
mm
a
r
y
(
R
O
U
G
E
-
1
=
1
,
R
O
U
G
E
-
2
=
1
)
Te
x
t
R
a
n
k
+
I
n
d
o
B
E
R
T
BA
S
E
+
TF
-
I
D
F
S
a
me
a
s t
h
e
s
u
mm
a
r
y
o
f
T
e
x
t
R
a
n
k
+
I
n
d
o
B
ER
T
BA
S
E
a
b
o
v
e
.
(
R
O
U
G
E
-
1
=
0
.
5
9
5
,
R
O
U
G
E
-
2
=
0
.
5
1
4
)
Te
x
t
R
a
n
k
+
I
n
d
o
B
E
R
T
L
A
RG
E
+
TF
-
I
D
F
(
En
g
l
i
s
h
t
r
a
n
sl
a
t
i
o
n
:
D
u
r
i
n
g
t
h
e
t
r
i
a
l
i
n
M
u
g
l
a
,
a
c
i
t
y
i
n
S
o
u
t
h
w
e
s
t
e
r
n
Tu
r
k
e
y
,
o
n
W
e
d
n
e
sd
a
y
(
1
0
/
0
4
)
,
t
h
e
j
u
d
g
e
s
t
a
t
e
d
t
h
a
t
t
h
e
d
e
f
e
n
d
a
n
t
s
a
t
t
a
c
k
e
d
a
h
o
t
e
l
i
n
t
h
e
sea
s
i
d
e
r
e
s
o
r
t
o
f
M
a
r
maris
,
w
h
e
r
e
P
r
e
s
i
d
e
n
t
Er
d
o
g
a
n
w
a
s
st
a
y
i
n
g
d
u
r
i
n
g
t
h
e
f
a
i
l
e
d
c
o
u
p
a
t
t
e
m
p
t
o
n
Ju
l
y
1
5
,
2
0
1
6
.
P
r
e
s
i
d
e
n
t
Er
d
o
g
a
n
h
i
m
sel
f
l
e
f
t
h
i
s
r
e
s
o
r
t
i
n
M
a
r
maris
b
e
f
o
r
e
t
h
e
a
t
t
a
c
k
,
w
h
i
c
h
k
i
l
l
e
d
t
w
o
p
o
l
i
c
e
o
f
f
i
c
e
r
s,
w
a
s
l
a
u
n
c
h
e
d
.
)
(
R
O
U
G
E
-
1
=
0
.
7
0
4
,
R
O
U
G
E
-
2
=
0
.
6
5
0
)
Te
x
t
R
a
n
k
+
W
o
r
d
2
V
e
c
(
I
n
d
o
S
u
m
)
+
TF
-
I
D
F
S
a
me
a
s t
h
e
g
r
o
u
n
d
t
r
u
t
h
su
mm
a
r
y
(
R
O
U
G
E
-
1
=
1
,
R
O
U
G
E
-
2
=
1
)
Te
x
t
R
a
n
k
+
F
a
st
Te
x
t
(
I
n
d
o
S
u
m
)
+
TF
-
I
D
F
S
a
me
a
s t
h
e
g
r
o
u
n
d
t
r
u
t
h
su
mm
a
r
y
(
R
O
U
G
E
-
1
=
1
,
R
O
U
G
E
-
2
=
1
)
Evaluation Warning : The document was created with Spire.PDF for Python.