I
nte
rna
t
io
na
l J
o
urna
l o
f
Adv
a
nces in Applie
d Science
s
(
I
J
AAS)
Vo
l.
14
,
No
.
2
,
J
u
n
e
2
0
2
5
,
p
p
.
562
~
5
7
9
I
SS
N:
2252
-
8
8
1
4
,
DOI
:
1
0
.
1
1
5
9
1
/ijaas
.
v14.
i
2
.
pp
562
-
5
7
9
562
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
a
s
.
ia
esco
r
e.
co
m
Sentimen
t
a
na
ly
sis
of va
ccine data
using
enhanced
d
eep
lea
rning
alg
o
rithms
M
o
nik
a
Ver
m
a
,
Sa
nd
ee
p M
o
ng
a
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
a
n
d
En
g
i
n
e
e
r
i
n
g
,
O
r
i
e
n
t
a
l
U
n
i
v
e
r
si
t
y
,
I
n
d
o
r
e
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Oct
1
5
,
2
0
2
4
R
ev
is
ed
Ap
r
25
,
2
0
2
5
Acc
ep
ted
May
10
,
2
0
2
5
Th
is
p
a
p
e
r
i
n
v
e
sti
g
a
tes
a
n
d
e
x
p
e
rime
n
ts
with
a
n
a
p
p
r
o
a
c
h
to
imp
ro
v
e
se
n
ti
m
e
n
t
a
n
a
ly
sis
o
n
v
a
c
c
in
e
d
a
tas
e
ts
with
d
e
e
p
lea
rn
i
n
g
.
I
t
e
v
a
lu
a
tes
ra
n
d
o
m
f
o
re
st
(RF
)
,
n
a
ïv
e
Ba
y
e
s
(NB)
,
a
n
d
re
c
u
rre
n
t
n
e
u
ra
l
n
e
tw
o
rk
(RNN
)
m
o
d
e
ls
a
c
ro
ss
a
v
a
riety
o
f
c
o
n
fi
g
u
ra
ti
o
n
s
,
i
.
e
.
,
v
e
c
t
o
r
d
ime
n
si
o
n
s,
p
o
o
li
n
g
tec
h
n
iq
u
e
s
,
a
s
we
ll
a
s
e
v
a
lu
a
ti
o
n
m
e
th
o
d
s
,
h
iera
rc
h
ica
l
S
o
f
tM
a
x
v
s
n
e
g
a
ti
v
e
sa
m
p
li
n
g
.
Th
e
re
su
lt
s
sh
o
w
t
h
a
t
th
e
m
o
d
e
l
we
p
r
o
p
o
se
d
p
re
v
a
il
e
d
wit
h
a
n
a
c
c
u
ra
c
y
o
f
9
9
.
0
5
%
o
n
a
lea
rn
i
n
g
ra
te
e
q
u
a
l
to
0
.
0
0
1
,
o
u
tp
e
rf
o
rm
in
g
a
ll
o
th
e
r
m
o
d
e
ls
b
a
se
d
o
n
m
e
tri
c
s
i
n
c
lu
d
in
g
p
re
c
isio
n
,
re
c
a
ll
,
a
n
d
F
1
-
sc
o
re
fo
r
b
e
n
ig
n
/
m
a
li
g
n
a
n
t
c
a
se
s.
Th
e
re
su
lt
s
su
g
g
e
st
th
a
t
h
i
g
h
e
r
v
e
c
t
o
r
d
ime
n
sio
n
s,
a
v
e
ra
g
e
p
o
o
li
n
g
,
lo
we
rin
g
th
e
d
ro
p
o
u
t
ra
te,
a
n
d
e
m
p
lo
y
in
g
h
i
e
ra
rc
h
ica
l
S
o
ftM
a
x
fo
r
o
u
tp
u
t
si
g
n
ifi
c
a
n
tl
y
imp
ro
v
e
m
o
d
e
l
p
e
rfo
rm
a
n
c
e
.
Hi
e
ra
rc
h
ica
l
S
o
ftM
a
x
p
e
rfo
rm
s
b
e
tt
e
r
t
h
a
n
n
e
g
a
ti
v
e
sa
m
p
li
n
g
,
wh
e
re
a
s
a
lo
we
r
d
ro
p
o
u
t
ra
te
d
e
c
re
a
s
e
s
o
v
e
rfit
ti
n
g
a
n
d
lea
d
s
to
imp
ro
v
e
d
g
e
n
e
ra
li
z
a
ti
o
n
.
O
u
r
re
su
l
ts
d
e
m
o
n
stra
te
th
e
n
e
c
e
ss
it
y
to
a
p
p
ly
m
o
re
so
p
h
isti
c
a
ted
d
e
e
p
-
lea
rn
in
g
t
o
o
ls
a
ro
u
n
d
c
a
p
tu
ri
n
g
n
u
a
n
c
e
s
o
f
p
u
b
l
ic
v
a
c
c
in
e
-
re
late
d
se
n
ti
m
e
n
t,
wh
ich
m
a
y
b
e
c
ru
c
ial
fo
r
i
n
fo
rm
i
n
g
c
o
m
m
u
n
i
c
a
ti
o
n
stra
teg
ies
a
n
d
su
p
p
o
rti
n
g
d
e
c
isio
n
-
m
a
k
in
g
in
a
re
a
l
-
wo
rld
h
e
a
lt
h
e
m
e
rg
e
n
c
y
.
Th
e
fin
d
in
g
s
in
d
ica
te
th
a
t
th
e
p
e
rfo
rm
a
n
c
e
o
f
se
n
ti
m
e
n
t
a
n
a
l
y
sis
with
re
g
a
rd
to
COV
ID
-
1
9
v
a
c
c
in
e
d
e
p
lo
y
m
e
n
t
p
o
li
c
y
d
e
sig
n
a
n
d
p
u
b
li
c
m
o
n
it
o
rin
g
c
o
u
l
d
b
e
e
n
h
a
n
c
e
d
b
y
a
d
v
a
n
c
e
d
d
e
e
p
lea
rn
in
g
a
l
g
o
rit
h
m
s.
K
ey
w
o
r
d
s
:
Bi
-
L
STM
m
o
d
el
Dee
p
lear
n
in
g
al
g
o
r
ith
m
s
Dr
o
p
o
u
t
r
ates
Hier
ar
ch
ical
So
f
tMa
x
Neg
ativ
e
s
am
p
lin
g
Sen
tim
en
t a
n
aly
s
is
Vac
cin
e
d
ataset
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
:
Mo
n
ik
a
Ver
m
a
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
an
d
E
n
g
in
ee
r
in
g
,
Or
ie
n
tal
Un
iv
er
s
ity
Gate
No
.
1
,
San
wer
R
d
,
o
p
p
o
s
i
te
R
ev
ati
R
an
g
e,
J
ak
h
y
a,
I
n
d
o
r
e,
Ma
d
h
y
a
Pra
d
esh
4
5
3
5
5
5
,
I
n
d
ia
E
m
ail: m
o
n
ik
av
e
r
m
a0
3
@
r
ed
if
f
.
co
m
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
r
ap
id
d
ev
elo
p
m
en
t
a
n
d
d
ep
lo
y
m
en
t
o
f
v
ac
cin
es,
p
ar
tic
u
lar
ly
in
r
esp
o
n
s
e
to
g
l
o
b
al
p
an
d
em
ics
lik
e
C
OVI
D
-
1
9
,
h
av
e
s
p
ar
k
e
d
ex
ten
s
iv
e
p
u
b
lic
d
is
co
u
r
s
e
an
d
v
ar
ied
s
en
tim
en
ts
ac
r
o
s
s
s
o
cial
m
ed
ia,
n
ews
p
latf
o
r
m
s
,
a
n
d
o
th
er
co
m
m
u
n
icatio
n
ch
an
n
els.
Un
d
e
r
s
tan
d
i
n
g
th
ese
s
en
tim
en
ts
is
cr
u
cial
f
o
r
p
u
b
lic
h
ea
lt
h
o
f
f
icials,
p
o
licy
m
a
k
er
s
,
an
d
r
esear
ch
er
s
to
g
a
u
g
e
p
u
b
lic
o
p
in
io
n
,
ad
d
r
ess
co
n
ce
r
n
s
,
a
n
d
cr
af
t
ef
f
ec
tiv
e
co
m
m
u
n
icatio
n
s
tr
ateg
ies.
Sen
tim
en
t
an
aly
s
is
,
a
f
ield
with
in
n
atu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
(
NL
P),
in
v
o
lv
es
d
eter
m
in
in
g
th
e
em
o
tio
n
al
t
o
n
e
b
e
h
in
d
tex
ts
an
d
is
in
cr
ea
s
in
g
ly
b
ein
g
ap
p
lie
d
to
s
t
u
d
y
v
ac
cin
e
-
r
elate
d
s
en
tim
en
ts
.
Ho
wev
er
,
an
al
y
zin
g
s
en
tim
en
ts
in
v
ac
cin
e
d
atasets
p
o
s
es
u
n
iq
u
e
c
h
a
llen
g
es
d
u
e
to
th
e
co
m
p
lex
ity
a
n
d
v
a
r
iab
ilit
y
o
f
th
e
tex
t
d
ata.
T
h
is
s
tu
d
y
f
o
c
u
s
es
o
n
en
h
an
cin
g
s
en
tim
en
t
an
aly
s
is
o
f
v
ac
cin
e
d
atasets
u
s
in
g
ad
v
an
ce
d
d
ee
p
lear
n
in
g
alg
o
r
ith
m
s
,
lev
er
ag
in
g
s
tate
-
of
-
th
e
-
ar
t
tech
n
i
q
u
es
s
u
ch
as
h
ier
ar
ch
ical
So
f
tMa
x
,
n
eg
ativ
e
s
am
p
lin
g
,
a
n
d
n
o
v
el
m
o
d
el
ar
ch
itectu
r
es t
o
im
p
r
o
v
e
ac
cu
r
ac
y
an
d
r
eliab
ilit
y
.
T
h
e
ef
f
ec
tiv
en
ess
o
f
d
ee
p
lear
n
in
g
m
o
d
els
in
s
en
tim
en
t
an
aly
s
is
is
s
ig
n
if
ican
tly
in
f
lu
en
ce
d
b
y
v
ar
io
u
s
f
ac
to
r
s
,
in
clu
d
i
n
g
v
ec
t
o
r
d
im
en
s
io
n
s
,
p
o
o
lin
g
tech
n
iq
u
es,
d
r
o
p
o
u
t
r
ates,
an
d
ev
al
u
atio
n
m
eth
o
d
s
.
Fo
r
ex
am
p
le,
in
c
r
ea
s
in
g
v
ec
to
r
d
im
en
s
io
n
s
f
r
o
m
1
0
0
to
3
0
0
h
as
b
ee
n
s
h
o
wn
to
en
h
a
n
ce
m
o
d
e
l
p
er
f
o
r
m
an
ce
,
as
it
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ad
v
Ap
p
l Sci
I
SS
N:
2252
-
8
8
1
4
S
en
timen
t a
n
a
lysi
s
o
f v
a
cc
in
e
d
a
ta
u
s
in
g
en
h
a
n
ce
d
d
ee
p
lea
r
n
in
g
a
l
g
o
r
ith
ms
(
Mo
n
ika
V
erma
)
563
allo
ws th
e
r
ep
r
esen
tatio
n
o
f
m
o
r
e
s
em
an
tic
n
u
an
ce
s
in
th
e
te
x
t.
Similar
ly
,
u
s
in
g
ad
v
a
n
ce
d
p
o
o
lin
g
tech
n
iq
u
es,
s
u
ch
as
av
er
ag
e
p
o
o
lin
g
o
v
er
m
ax
p
o
o
lin
g
,
ca
n
r
esu
lt
in
b
e
tter
m
o
d
el
ac
c
u
r
ac
y
b
y
e
f
f
ec
ti
v
ely
ca
p
tu
r
in
g
t
h
e
m
o
s
t
r
ep
r
esen
tativ
e
f
ea
tu
r
es
f
r
o
m
th
e
d
ata.
T
h
is
s
tu
d
y
also
h
ig
h
lig
h
ts
th
e
im
p
o
r
tan
ce
o
f
f
in
e
-
tu
n
in
g
m
o
d
el
p
ar
am
eter
s
,
s
u
ch
as
d
r
o
p
o
u
t
v
alu
es
an
d
lear
n
in
g
r
ates,
to
p
r
ev
en
t
o
v
e
r
f
itti
n
g
an
d
im
p
r
o
v
e
g
en
er
aliza
tio
n
o
n
u
n
s
ee
n
d
ata.
T
h
is
r
esear
ch
d
em
o
n
s
tr
ates
th
e
p
o
te
n
tial
o
f
ad
v
a
n
ce
d
d
ee
p
-
lear
n
in
g
alg
o
r
ith
m
s
f
o
r
en
h
an
cin
g
s
en
tim
en
t
an
aly
s
is
o
f
v
ac
ci
n
e
d
atasets
.
B
y
lev
er
a
g
in
g
s
tate
-
of
-
th
e
-
ar
t
tec
h
n
iq
u
es
s
u
ch
as
h
ie
r
ar
ch
ical
So
f
tMa
x
,
av
er
ag
e
p
o
o
lin
g
,
a
n
d
o
p
tim
ized
p
a
r
am
eter
s
ettin
g
s
,
th
e
p
r
o
p
o
s
ed
m
o
d
el
ac
h
iev
es
u
n
p
r
ec
ed
e
n
ted
ac
cu
r
ac
y
an
d
r
o
b
u
s
tn
ess
.
T
h
ese
f
in
d
in
g
s
n
o
t
o
n
l
y
co
n
tr
i
b
u
te
to
th
e
f
ield
o
f
s
en
tim
e
n
t
an
aly
s
is
b
u
t
also
p
r
o
v
id
e
v
al
u
ab
le
i
n
s
ig
h
ts
f
o
r
p
u
b
lic
h
ea
lth
c
o
m
m
u
n
icatio
n
an
d
p
o
licy
-
m
a
k
in
g
,
p
a
r
ticu
lar
l
y
in
u
n
d
er
s
tan
d
in
g
an
d
r
esp
o
n
d
in
g
to
p
u
b
lic
s
en
tim
en
ts
ab
o
u
t
v
ac
cin
es.
F
u
t
u
r
e
w
o
r
k
c
a
n
e
x
p
l
o
r
e
t
h
e
i
n
t
e
g
r
a
t
i
o
n
o
f
o
t
h
e
r
d
e
e
p
l
e
a
r
n
i
n
g
a
r
c
h
i
t
ec
t
u
r
e
s
,
s
u
c
h
as
t
r
a
n
s
f
o
r
m
e
r
s
,
a
n
d
e
x
t
e
n
d
t
h
e
an
a
l
y
s
is
t
o
m
u
lt
i
li
n
g
u
a
l
v
a
c
c
i
n
e
d
a
t
a
s
et
s
t
o
f
u
r
t
h
e
r
e
n
h
a
n
c
e
t
h
e
g
e
n
e
r
a
l
iz
a
b
i
l
it
y
an
d
i
m
p
a
c
t
o
f
t
h
e
s
e
m
o
d
e
l
s.
T
h
e
c
o
n
t
r
i
b
u
t
i
o
n
s
o
f
t
h
i
s
s
t
u
d
y
a
r
e
s
u
m
m
a
r
is
e
d
as
f
o
l
l
o
w
s
:
i
)
W
e
p
r
o
p
o
s
e
a
n
in
n
o
v
ativ
e
d
ee
p
-
lea
r
n
in
g
m
o
d
el
f
o
r
s
en
tim
en
t
a
n
aly
s
is
,
s
p
ec
if
ically
d
esig
n
ed
f
o
r
v
ac
cin
e
-
r
elate
d
d
atasets
.
T
h
e
m
o
d
el
in
teg
r
ates
b
id
ir
ec
tio
n
al
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
Bi
-
L
STM
)
with
atten
tio
n
m
ec
h
an
is
m
s
,
o
p
tim
izin
g
s
en
tim
en
t
class
if
icatio
n
b
y
ca
p
tu
r
in
g
co
n
tex
tu
al
in
f
o
r
m
atio
n
f
r
o
m
th
e
tex
t
;
ii)
Hier
ar
ch
ical
So
f
tMa
x
is
em
p
lo
y
ed
as
a
s
u
p
e
r
io
r
ev
al
u
atio
n
m
eth
o
d
o
v
e
r
n
eg
ativ
e
s
a
m
p
lin
g
,
s
ig
n
if
ican
tly
im
p
r
o
v
in
g
m
o
d
el
p
er
f
o
r
m
an
ce
an
d
ac
cu
r
ac
y
;
iii)
W
e
s
y
s
tem
atica
lly
in
v
esti
g
ate
th
e
im
p
ac
t
o
f
v
ar
i
o
u
s
co
n
f
ig
u
r
atio
n
s
,
in
cl
u
d
in
g
v
ec
t
o
r
d
im
e
n
s
io
n
s
,
p
o
o
lin
g
tech
n
i
q
u
es
(
av
er
a
g
e
p
o
o
lin
g
v
e
r
s
u
s
m
ax
p
o
o
lin
g
)
,
an
d
d
r
o
p
o
u
t
r
ates,
r
e
v
ea
lin
g
th
e
ir
in
f
lu
en
ce
o
n
m
o
d
el
ac
c
u
r
ac
y
an
d
g
e
n
er
aliza
tio
n
;
a
n
d
iv
)
T
h
e
s
tu
d
y
d
em
o
n
s
tr
ates
th
at
h
ig
h
er
-
d
im
en
s
io
n
al
em
b
ed
d
in
g
s
an
d
lo
wer
d
r
o
p
o
u
t
r
ates
en
h
an
ce
th
e
m
o
d
el’
s
ab
ilit
y
to
lear
n
co
m
p
lex
s
en
tim
en
t
p
att
er
n
s
an
d
p
r
e
v
en
t
o
v
er
f
itti
n
g
,
o
f
f
er
in
g
cr
u
cial
in
s
ig
h
ts
f
o
r
i
m
p
r
o
v
i
n
g
s
en
tim
en
t
an
aly
s
is
in
p
u
b
lic
h
ea
lth
c
o
m
m
u
n
icatio
n
.
T
h
e
r
est
o
f
th
e
p
ap
er
is
d
i
v
id
ed
in
to
f
iv
e
s
ec
tio
n
s
:
s
ec
tio
n
2
p
r
esen
ts
a
liter
atu
r
e
r
ev
ie
w
o
f
ex
is
tin
g
s
en
tim
en
t
an
aly
s
is
tech
n
iq
u
es
f
o
r
v
ac
cin
e
d
atasets
;
s
ec
tio
n
3
d
etails
th
e
p
r
o
p
o
s
ed
m
eth
o
d
o
lo
g
y
,
i
n
clu
d
in
g
th
e
ad
v
an
ce
d
d
ee
p
lear
n
i
n
g
al
g
o
r
i
th
m
s
em
p
lo
y
e
d
;
s
ec
tio
n
4
d
is
cu
s
s
es
th
e
r
esu
lts
,
p
r
o
v
id
in
g
a
n
in
-
d
ep
th
an
aly
s
is
o
f
th
e
f
i
n
d
in
g
s
an
d
th
ei
r
im
p
licatio
n
s
; a
n
d
s
ec
tio
n
5
c
o
n
clu
d
es th
e
s
tu
d
y
,
s
u
m
m
ar
izin
g
th
e
k
ey
o
u
tc
o
m
es.
2.
L
I
T
E
R
AT
U
RE
R
E
VI
E
W
So
cial
m
ed
ia
an
d
o
n
lin
e
f
o
r
u
m
s
allo
w
in
d
iv
id
u
als
to
d
e
b
at
e
p
u
b
lic
h
ea
lth
t
o
p
ics
lik
e
C
OVI
D
-
1
9
a
n
d
s
p
r
ea
d
co
r
r
ec
t
an
d
in
ac
c
u
r
a
te
in
f
o
r
m
atio
n
.
T
h
is
s
tu
d
y
in
t
r
o
d
u
ce
s
B
i
-
L
STM
-
b
ased
NL
P.
C
lass
if
y
in
g
s
en
tim
en
ts
an
d
id
en
tify
in
g
is
s
u
es
r
elate
d
to
C
O
VI
D
-
1
9
p
u
b
lic
co
m
m
en
ts
is
t
h
e
g
o
al.
T
o
im
p
r
o
v
e
tr
a
d
itio
n
al
L
STM
s
,
B
i
-
L
STM
u
s
e
s
o
u
tp
u
ts
f
r
o
m
p
r
ev
io
u
s
an
d
s
u
b
s
eq
u
en
t
d
ata
co
n
te
x
ts
.
O
u
r
p
e
r
f
o
r
m
a
n
c
e
m
e
t
r
i
c
s
e
x
c
e
e
d
e
d
t
r
a
d
i
t
i
o
n
a
l
L
S
T
M
m
o
d
e
l
s
a
n
d
p
r
i
o
r
r
es
e
a
r
c
h
a
f
t
e
r
a
n
a
l
y
z
i
n
g
T
wi
t
t
e
r
a
n
d
R
e
d
d
i
t
d
a
t
a
s
et
s
.
T
h
i
s
n
o
t
i
o
n
h
e
l
p
s
g
o
v
e
r
n
m
e
n
t
s
r
e
d
u
c
e
n
eg
a
t
i
v
e
c
o
m
m
u
n
i
c
a
ti
o
n
a
n
d
u
n
d
e
r
s
t
a
n
d
p
u
b
l
i
c
o
p
i
n
i
o
n
d
u
r
i
n
g
h
e
a
l
t
h
c
r
is
es
.
O
u
r
w
o
r
k
a
l
s
o
u
n
d
e
r
l
i
n
es
t
h
e
n
e
e
d
to
e
m
p
l
o
y
N
L
P
t
o
a
n
a
l
y
z
e
p
u
b
l
ic
m
o
o
d
,
w
h
i
c
h
m
a
y
g
u
i
d
e
h
e
a
l
th
p
o
l
i
c
y
[
1
]
.
Af
ter
th
e
2
0
1
9
C
OVI
D
-
1
9
p
a
n
d
em
ic,
wh
ich
is
ass
u
m
ed
to
h
av
e
s
tar
ted
i
n
W
u
h
a
n
,
C
h
i
n
a,
p
e
o
p
le
g
lo
b
ally
p
r
ac
ticed
h
an
d
h
y
g
ien
e,
f
ac
e
m
ask
s
,
a
n
d
p
h
y
s
ical
d
is
tan
cin
g
.
C
OVI
D
-
1
9
v
ac
cin
atio
n
s
wer
e
in
tr
o
d
u
ce
d
in
ea
r
ly
2
0
2
1
in
s
ev
er
al
co
u
n
tr
ies,
in
clu
d
in
g
th
e
US,
b
r
i
n
g
in
g
r
elief
b
u
t
also
p
o
lar
izin
g
d
eb
ate.
Vac
cin
e
r
elu
ctan
ce
b
ec
a
m
e
a
m
ajo
r
p
r
o
b
lem
af
ter
th
ese
c
o
n
v
er
s
atio
n
s
.
Sin
ce
c
o
n
v
e
n
tio
n
al
d
ata
is
lim
ited
,
liv
e
-
s
tr
ea
m
ed
twee
ts
f
r
o
m
API
q
u
er
ies
p
r
o
v
i
d
e
a
v
iab
le
way
to
s
tu
d
y
p
u
b
lic
o
p
in
i
o
n
o
n
v
a
cc
in
e
is
s
u
es.
Azu
r
e
ma
ch
in
e
lear
n
in
g
(
ML
)
,
VA
DE
R
,
an
d
T
e
x
tB
lo
b
wer
e
e
m
p
lo
y
ed
in
th
is
wo
r
k
to
g
eth
er
with
f
i
v
e
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
an
d
th
r
ee
tex
t
v
ec
to
r
i
z
atio
n
m
eth
o
d
s
f
o
r
s
en
tim
en
t
an
aly
s
is
.
Ou
r
co
m
p
r
eh
en
s
iv
e
m
o
d
el
ev
alu
atio
n
s
h
o
wed
p
u
b
lic
s
en
tim
en
t
o
n
C
OVI
D
-
1
9
im
m
u
n
i
z
at
io
n
is
im
p
r
o
v
in
g
.
T
h
e
b
es
t
p
u
b
lic
s
en
tim
en
t
p
r
ed
ictio
n
m
etr
ics
wer
e
ac
h
iev
ed
u
s
in
g
T
ex
tB
lo
b
s
en
tim
en
t
s
co
r
e,
ter
m
f
r
eq
u
en
cy
-
in
v
er
s
e
d
o
cu
m
en
t
f
r
eq
u
e
n
cy
(
TF
-
I
DF
)
v
ec
to
r
i
z
at
io
n
,
an
d
L
in
ea
r
SVC
class
if
icat
io
n
[
2
]
.
Sear
ch
en
g
in
es
an
d
s
o
cial
n
etwo
r
k
s
h
av
e
b
ee
n
u
s
ed
in
in
ter
n
et
-
b
ased
s
y
n
d
r
o
m
ic
m
o
n
ito
r
in
g
to
p
r
ed
ict
ep
id
em
ic
ten
d
e
n
cies
f
o
r
2
0
y
ea
r
s
.
R
ec
en
tly
,
th
e
f
o
c
u
s
h
as
s
h
if
ted
to
u
n
d
er
s
tan
d
in
g
p
u
b
lic
em
o
tio
n
al
r
esp
o
n
s
es
to
h
ea
lth
c
r
is
es,
n
o
tab
ly
p
an
d
em
ics.
T
h
is
s
tu
d
y
e
x
am
in
ed
Gr
ee
k
T
witter
's
attitu
d
e
to
war
d
s
C
OVI
D
-
1
9
ca
s
es
in
r
ea
l
-
tim
e.
T
wo
em
o
tio
n
lex
ico
n
s
o
n
e
tr
an
s
lated
f
r
o
m
E
n
g
lis
h
to
Gr
ee
k
an
d
o
n
e
o
r
ig
in
al
Gr
ee
k
wer
e
u
s
ed
to
an
aly
z
e
1
5
3
,
5
2
8
twee
ts
f
r
o
m
1
8
,
7
3
0
p
e
r
s
o
n
s
.
T
h
e
s
tu
d
y
m
ea
s
u
r
ed
p
o
s
itiv
e
an
d
n
eg
ativ
e
att
itu
d
es
an
d
s
ix
em
o
tio
n
s
.
W
e
a
ls
o
e
x
a
m
i
n
e
d
a
t
ti
t
u
d
e
s
,
t
w
e
e
t
v
o
l
u
m
e
s
,
a
n
d
C
OV
I
D
-
1
9
c
a
s
es
.
T
h
e
m
o
s
t
c
o
m
m
o
n
e
m
o
t
i
o
n
s
w
e
r
e
s
u
r
p
r
is
e
(
2
5
.
3
2
%
)
a
n
d
d
is
g
u
s
t
(
1
9
.
8
8
%
)
.
H
o
we
v
e
r
,
m
o
o
d
d
i
d
n
o
t
s
ig
n
i
f
i
c
a
n
t
l
y
c
o
r
r
e
la
t
e
w
i
t
h
C
O
V
I
D
-
1
9
d
i
s
s
e
m
i
n
a
ti
o
n
,
i
n
d
i
c
a
t
i
n
g
t
h
at
i
n
t
e
r
es
t
m
a
y
d
ec
l
i
n
e
w
it
h
t
i
m
e
[
3
]
.
I
n
d
o
n
esia'
s
C
OVI
D
-
1
9
an
s
w
er
s
p
lit
T
witter
u
s
er
s
.
T
h
e
e
x
am
in
atio
n
o
f
th
ese
twee
ts
m
ay
in
f
o
r
m
p
o
licy
m
ak
in
g
an
d
g
o
v
er
n
m
e
n
t
ass
ess
m
en
t.
Sen
tim
en
t
an
aly
s
is
u
s
es
twee
ts
to
as
s
ess
p
u
b
lic
o
p
in
io
n
.
T
h
is
r
esear
ch
ex
am
in
es
p
u
b
lic
o
p
i
n
io
n
o
n
I
n
d
o
n
esia
'
s
ep
id
em
ic
m
an
ag
em
en
t
f
r
o
m
g
e
n
er
al
an
d
ec
o
n
o
m
ic
asp
ec
ts
.
T
h
e
T
witter
s
cr
ap
er
lib
r
ar
y
co
l
lects
twee
t
s
.
Sen
tis
tr
en
g
th
_
id
an
d
ex
p
e
r
t
ass
ess
m
en
t
clas
s
if
i
ed
th
ese
twee
ts
as
f
av
o
r
a
b
le,
n
eg
ativ
e,
o
r
n
e
u
tr
al.
Data
p
r
e
-
p
r
o
ce
s
s
in
g
r
em
o
v
ed
ex
tr
an
e
o
u
s
twee
ts
.
C
o
n
f
u
s
io
n
m
atr
ices
an
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
8
1
4
I
n
t J Ad
v
Ap
p
l Sci
,
Vo
l.
14
,
No
.
2
,
J
u
n
e
2
0
2
5
:
5
6
2
-
579
564
K
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
c
o
n
f
ir
m
ed
th
e
ac
cu
r
ac
y
o
f
m
ac
h
in
e
lear
n
in
g
m
o
d
els
th
at
ass
es
s
ed
f
r
esh
d
ata
s
en
tim
en
t.
S
u
p
p
o
r
t
v
e
cto
r
m
ac
h
in
e
(
SVM
)
r
esear
ch
y
iel
d
ed
im
p
r
ess
iv
e
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
an
d
f
-
m
ea
s
u
r
e
s
co
r
es
o
f
8
2
.
0
0
,
8
2
.
2
4
,
8
2
.
0
1
,
an
d
8
1
.
8
4
%.
C
iv
ilian
s
ten
d
ed
to
lik
e
th
e
ep
id
em
ic'
s
ec
o
n
o
m
ic
p
o
licies,
b
u
t
m
an
y
wer
e
d
is
s
atis
f
ied
with
th
e
g
o
v
er
n
m
e
n
t'
s
p
er
f
o
r
m
an
ce
.
T
h
e
SVM
ap
p
r
o
ac
h
,
n
o
tab
l
y
with
th
e
No
r
m
alize
d
Po
ly
Ke
r
n
el,
p
r
ed
icts
T
witter
s
en
tim
en
t r
ap
id
ly
an
d
ac
cu
r
ately
[
4
]
.
T
h
e
C
OVI
D
-
1
9
p
an
d
em
ic
ca
u
s
ed
h
ea
lth
an
d
ec
o
n
o
m
ic
p
r
o
b
lem
s
.
B
ig
d
ata
h
elp
s
lo
g
is
tics
b
u
s
in
ess
es
f
in
d
p
r
o
f
itab
le
s
o
lu
tio
n
s
to
th
ese
d
if
f
icu
lt
s
itu
atio
n
s
.
T
h
is
s
tu
d
y
an
aly
z
ed
lo
g
is
tics
co
n
s
u
ltan
cy
web
s
ites
u
s
in
g
tex
t
m
in
in
g
.
T
h
e
m
ai
n
o
b
jec
ti
v
es
w
er
e
t
o
d
et
ec
t
f
r
e
q
u
e
n
t
p
h
r
as
es,
u
n
d
er
ta
k
e
s
en
ti
m
e
n
t
a
n
a
ly
s
is
u
s
i
n
g
t
h
e
NR
C
lex
ic
o
n
,
s
t
r
ess
c
o
m
m
o
n
w
o
r
d
c
o
m
b
i
n
ati
o
n
s
,
a
n
d
p
r
o
v
i
d
e
c
o
s
t
-
e
f
f
ec
ti
v
e
s
h
ip
p
i
n
g
a
n
d
i
n
v
en
t
o
r
y
m
a
n
a
g
em
en
t
s
o
l
u
ti
o
n
s
th
r
o
u
g
h
o
u
t
t
h
e
p
a
n
d
em
i
c.
I
m
p
o
r
t
a
n
t
p
h
r
ases
w
er
e
"su
p
p
l
y
,
"
"c
h
ai
n
,
"
a
n
d
"C
OVI
D
-
1
9
.
"
T
r
u
s
t
-
r
elate
d
em
o
tio
n
s
wer
e
p
r
esen
t,
s
h
o
win
g
a
d
esire
f
o
r
r
eliab
le
r
em
ed
i
es th
r
o
u
g
h
o
u
t th
e
cr
is
is
[
5
]
.
I
n
f
o
r
m
atio
n
o
n
s
o
cial
m
e
d
ia,
p
ar
ticu
lar
l
y
o
n
t
h
e
C
OVI
D
-
1
9
p
a
n
d
em
ic
,
is
f
r
eq
u
e
n
tly
f
alse.
T
h
is
r
esear
ch
an
aly
s
es
Ma
lay
s
ian
C
OVI
D
-
1
9
n
ews
s
en
tim
en
t
o
n
T
witter
.
W
e
co
llected
lar
g
e
ly
Ma
lay
,
E
n
g
lis
h
,
an
d
C
h
in
ese
twee
ts
s
in
ce
Ma
l
ay
s
ian
twee
ts
ar
e
m
u
ltil
in
g
u
al
.
Af
ter
a
n
aly
z
in
g
th
e
b
y
te
-
p
ai
r
en
co
d
in
g
-
tex
t
-
to
-
im
ag
e
-
co
n
v
o
lu
t
io
n
al
n
e
u
r
al
n
etwo
r
k
(
B
PE
-
T
ex
t
-
to
-
I
m
a
g
e
-
C
NN
)
an
d
B
PE
-
m
u
ltil
in
g
u
al
b
id
ir
ec
tio
n
al
en
co
d
er
r
ep
r
esen
tatio
n
s
f
r
o
m
tr
an
s
f
o
r
m
er
s
(
B
PE
-
M
-
B
E
R
T
)
m
o
d
els,
we
ch
o
s
e
th
e
p
r
e
-
tr
ain
ed
M
-
B
E
R
T
n
etwo
r
k
f
o
r
o
u
r
m
u
ltil
in
g
u
al
d
ataset
[
6
]
.
T
witter
h
as
m
ad
e
it
ea
s
ier
to
f
o
llo
w
g
lo
b
al
ev
e
n
ts
an
d
p
u
b
lic
v
iewp
o
in
ts
,
esp
ec
ially
am
id
h
ea
lth
cr
is
es.
T
h
e
C
O
VI
D
-
1
9
an
d
MPo
x
p
an
d
em
ics
b
o
o
s
ted
T
witter
co
m
m
u
n
icatio
n
.
T
h
is
s
tu
d
y
an
aly
s
es
twee
ts
ab
o
u
t
two
c
o
n
d
itio
n
s
s
im
u
ltan
eo
u
s
ly
to
f
ill
a
r
ese
ar
ch
g
a
p
.
Ma
in
ly
n
e
g
ativ
e
o
p
in
io
n
s
an
d
n
u
m
er
o
u
s
allu
s
io
n
s
to
Pre
s
id
en
t Bi
d
en
an
d
Uk
r
ain
e
g
iv
e
a
co
m
p
lete
u
n
d
er
s
tan
d
i
n
g
o
f
p
o
p
u
lar
d
is
co
u
r
s
e
in
th
is
ag
e
[
7
]
.
C
OVI
D
-
1
9
p
r
ev
en
tio
n
m
ea
s
u
r
es
wer
e
im
p
lem
en
ted
,
b
u
t
th
e
co
m
m
u
n
ity
'
s
ig
n
o
r
a
n
ce
an
d
lack
o
f
s
elf
-
co
n
tr
o
l
r
e
n
d
er
e
d
th
em
in
e
f
f
ec
tiv
e.
T
h
is
o
v
er
s
ig
h
t
in
c
r
ea
s
ed
v
ir
al
ex
p
o
s
u
r
e.
I
n
ad
d
it
io
n
to
h
ea
lth
,
th
e
p
an
d
em
ic
h
a
d
m
ajo
r
im
p
ac
ts
o
n
s
o
cial,
p
o
lit
ical,
r
elig
io
u
s
,
ec
o
n
o
m
ic,
a
n
d
p
o
p
u
latio
n
r
esil
ien
ce
.
So
cial
m
ed
ia
also
s
h
ed
s
lig
h
t
o
n
th
ese
ef
f
e
cts,
esp
ec
ially
s
o
cio
ec
o
n
o
m
ic
o
n
es.
T
h
is
s
tu
d
y
u
s
es
T
witter
API
d
ata
to
ass
ess
I
n
d
o
n
esian
s
'
p
an
d
em
ic
v
iews.
W
e
cr
ea
ted
a
s
en
tim
en
t
an
aly
s
is
to
o
l
u
s
in
g
T
F
-
I
DF,
L
e
x
ical,
an
d
NB
f
o
r
f
ea
tu
r
e
ex
tr
ac
tio
n
a
n
d
class
if
icatio
n
.
I
n
d
o
n
esian
T
witter
twee
ts
o
n
"COVI
D
-
1
9
"
wer
e
class
ed
b
y
em
o
tio
n
:
f
ea
r
,
r
ag
e,
lo
v
e,
g
r
ief
,
a
n
d
h
ap
p
in
ess
.
Ou
r
p
r
o
p
o
s
ed
tr
a
n
s
cr
ip
tio
n
f
ac
to
r
b
i
n
d
in
g
s
ite
(
T
FB
S
)
alg
o
r
ith
m
o
u
tp
er
f
o
r
m
ed
o
th
er
s
with
an
ac
cu
r
ac
y
o
f
0
.
8
5
.
Acc
u
r
ac
y
,
r
ec
all,
an
d
F
-
s
co
r
e
d
em
o
n
s
tr
at
ed
th
eir
s
u
p
er
io
r
ity
[
8
]
.
Sen
tim
en
t
an
aly
s
is
o
f
Gr
ee
k
Natio
n
al
Pu
b
lic
Hea
lth
O
r
g
an
izatio
n
(
E
ODY)
Face
b
o
o
k
p
o
s
ts
d
u
r
in
g
th
e
p
an
d
em
ic
was
p
er
f
o
r
m
ed
u
s
in
g
Mic
r
o
s
o
f
t
Azu
r
e
Ma
ch
i
n
e
L
ea
r
n
in
g
Stu
d
i
o
.
An
ex
a
m
in
a
tio
n
o
f
3
0
0
r
e
v
iews
id
en
tifie
d
o
p
in
i
o
n
s
as
p
o
s
itiv
e,
n
eg
ativ
e,
o
r
n
eu
tr
al.
T
h
e
r
e
s
ea
r
ch
ex
am
in
ed
r
ea
cti
o
n
s
to
E
ODY'
s
Face
b
o
o
k
d
aily
C
OVI
D
-
1
9
s
u
r
v
eillan
ce
r
ep
o
r
ts
.
T
h
e
s
en
tim
en
t
class
if
icatio
n
s
wer
e
au
to
m
ated
u
s
in
g
m
ac
h
in
e
lear
n
i
n
g
.
Go
v
er
n
m
e
n
t,
im
m
u
n
i
z
atio
n
s
,
an
d
C
OVI
D
-
1
9
wer
e
o
f
ten
m
en
tio
n
ed
,
in
d
icatin
g
d
is
ap
p
r
o
v
al
o
f
th
ese
r
ep
o
r
ts
.
Neu
r
al
n
etwo
r
k
(
NN)
an
d
NB
p
o
in
t
m
ac
h
in
e
class
if
icatio
n
m
eth
o
d
s
h
av
e
h
i
g
h
ac
cu
r
ac
y
an
d
F1
-
s
co
r
es.
I
n
p
an
d
em
ics,
m
a
ch
in
e
lear
n
in
g
ca
n
m
ea
s
u
r
e
p
u
b
lic
o
p
i
n
io
n
an
d
aid
p
u
b
li
c
h
ea
lth
d
ec
is
io
n
-
m
ak
in
g
[
9
]
.
Palliativ
e
ca
r
e
b
ec
am
e
in
cr
ea
s
in
g
ly
im
p
o
r
ta
n
t
as
th
e
d
is
ea
s
e
s
p
r
ea
d
.
T
h
is
s
tu
d
y
in
v
esti
g
ated
o
v
e
r
2
6
,
0
0
0
E
n
g
lis
h
twee
ts
f
r
o
m
2
0
2
0
to
2
0
2
2
to
d
eter
m
in
e
p
alliativ
e
ca
r
e
v
iews
d
u
r
in
g
th
e
p
an
d
em
ic.
Fo
u
r
th
em
es
em
er
g
ed
f
r
o
m
web
s
cr
ap
in
g
.
No
ta
b
ly
,
alth
o
u
g
h
m
an
y
twee
ts
h
ig
h
lig
h
ted
th
e
p
an
d
em
ic'
s
n
eg
ativ
e
ef
f
ec
ts
o
n
p
alliativ
e
ca
r
e,
m
an
y
also
h
ig
h
lig
h
ted
its
p
o
s
itiv
e
ef
f
ec
ts
.
Ma
ch
in
e
lear
n
in
g
ca
te
g
o
r
i
z
ed
s
ev
er
al
o
f
th
ese
twee
ts
,
f
o
cu
s
in
g
o
n
C
OVI
D
-
1
9
'
s
n
eg
ativ
e
ef
f
ec
ts
[
1
0
]
.
T
witter
's
wid
e
p
u
b
lic
o
p
in
io
n
m
ay
b
r
in
g
f
r
esh
an
d
ess
en
tial in
f
o
r
m
atio
n
,
esp
ec
ially
o
n
v
ac
cin
e
h
esit
an
cy
.
T
h
e
cu
r
r
e
n
t r
esear
ch
p
r
ep
r
o
ce
s
s
ed
an
d
ca
teg
o
r
i
z
ed
s
u
b
ject
-
r
elate
d
twee
ts
b
y
v
ar
i
o
u
s
attitu
d
es
an
d
em
o
tio
n
s
u
s
in
g
NR
C
L
ex
ico
n
.
Statis
tica
l
test
in
g
co
n
f
ir
m
ed
em
o
tio
n
al
co
r
r
elatio
n
s
.
Af
te
r
tr
ain
in
g
m
an
y
n
e
u
r
al
n
etwo
r
k
s
f
o
r
s
en
tim
en
t
m
u
lti
-
class
if
icatio
n
,
th
e
B
E
R
T
m
o
d
el
o
b
tain
ed
9
6
.
7
1
% a
cc
u
r
ac
y
[
1
1
]
.
T
h
is
s
tu
d
y
an
aly
s
es
T
witter
d
ata
to
ex
am
in
e
e
p
id
em
ic
-
r
elat
ed
m
o
b
ilit
y
m
o
d
e
p
r
ef
er
e
n
ce
s
.
J
an
u
ar
y
2020
to
J
an
u
ar
y
2
0
2
2
twee
ts
co
n
ce
r
n
in
g
New
Y
o
r
k
C
ity
tr
an
s
it
m
o
d
al
ities
wer
e
g
ath
e
r
e
d
.
W
e
cr
ea
ted
NL
P
-
b
ased
tr
av
el
m
o
d
e
class
if
ier
s
to
ca
teg
o
r
i
z
e
twee
ts
in
to
m
u
lt
ip
le
tr
an
s
p
o
r
tatio
n
m
o
d
es
u
s
in
g
th
is
d
ata.
Du
r
in
g
th
e
o
u
t
b
r
ea
k
,
p
u
b
lic
s
en
tim
en
t
ch
an
g
ed
t
o
war
d
s
tr
a
n
s
p
o
r
tati
o
n
c
h
o
ices.
B
u
s
es,
b
icy
cles,
a
n
d
p
r
iv
ate
ca
r
s
wer
e
well
-
r
eg
ar
d
ed
s
in
ce
co
m
m
u
ter
s
ch
o
s
e
th
em
o
v
er
s
u
b
way
s
.
T
witter
p
o
s
ts
o
n
th
e
tu
b
e
an
d
b
u
s
m
ask
v
io
latio
n
s
r
aised
co
n
ce
r
n
s
.
A
r
eg
r
ess
io
n
s
tu
d
y
em
p
lo
y
in
g
u
s
er
d
em
o
g
r
ap
h
ics
in
d
icate
d
th
e
f
ac
to
r
s
th
at
in
f
lu
en
ce
p
u
b
lic
tr
an
s
p
o
r
t
attitu
d
es,
p
ar
ticu
lar
ly
th
e
s
er
v
i
ce
s
ec
to
r
'
s
v
u
ln
e
r
ab
ilit
y
to
m
etr
o
p
o
litan
t
r
an
s
p
o
r
tatio
n
au
th
o
r
it
y
(
MTA
)
s
u
b
way
p
e
r
f
o
r
m
an
ce
[
1
2
]
.
C
OVI
D
-
1
9
'
s
g
lo
b
al
p
an
d
em
i
c
ca
u
s
ed
a
n
x
iety
an
d
d
r
o
v
e
m
an
y
So
u
th
Af
r
ican
s
to
r
eli
g
io
u
s
r
ites
.
So
cial
m
ed
ia
d
ata
is
an
aly
z
ed
to
d
eter
m
in
e
So
u
th
Af
r
ica
n
s
'
v
iews
o
n
r
elig
i
o
n
a
n
d
well
-
b
e
in
g
th
r
o
u
g
h
o
u
t
th
e
p
an
d
em
ic.
W
e
an
aly
z
e
d
twee
ts
ab
o
u
t
C
OVI
D
-
1
9
,
r
elig
io
n
,
life
'
s
p
u
r
p
o
s
e,
an
d
p
e
r
s
o
n
al
ex
p
er
ien
ce
s
to
d
eter
m
in
e
s
en
tim
en
t.
T
h
is
r
es
ea
r
ch
s
h
o
ws
th
at
r
elig
i
o
u
s
b
e
lief
s
an
d
C
OVI
D
-
1
9
attitu
d
es
af
f
ec
t
life
e
v
en
ts
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ad
v
Ap
p
l Sci
I
SS
N:
2252
-
8
8
1
4
S
en
timen
t a
n
a
lysi
s
o
f v
a
cc
in
e
d
a
ta
u
s
in
g
en
h
a
n
ce
d
d
ee
p
lea
r
n
in
g
a
l
g
o
r
ith
ms
(
Mo
n
ika
V
erma
)
565
Al
s
o
,
we
p
r
o
p
o
s
e
a
n
o
v
el
s
en
tim
en
t
an
aly
s
is
th
r
esh
o
ld
o
f
d
ep
r
ess
io
n
m
ea
s
u
r
e
th
at
p
r
o
v
i
d
es
u
s
ef
u
l
in
s
ig
h
ts
in
to
co
llectiv
e
em
o
tio
n
al
ci
r
cu
m
s
tan
ce
s
d
u
r
in
g
c
r
is
es
[
1
3
]
.
T
witter
b
ec
am
e
a
g
lo
b
al
c
o
n
v
er
s
atio
n
v
en
u
e
d
u
r
in
g
th
e
C
OVI
D
-
1
9
p
an
d
e
m
ic.
T
h
er
e
is
li
ttle
s
tu
d
y
o
n
th
e
p
o
ten
tial
u
s
ef
u
ln
ess
o
f
s
en
tim
en
t
an
aly
s
is
o
f
ep
id
em
i
c
twee
ts
,
d
esp
ite
v
ar
io
u
s
s
tu
d
ies
o
n
th
is
p
latf
o
r
m
.
T
h
is
r
esear
ch
s
h
o
ws
h
o
w
s
en
t
im
en
t
an
aly
s
is
p
air
ed
with
b
e
h
av
io
r
al
an
d
s
o
cial
s
cien
ce
m
ig
h
t
h
elp
p
an
d
em
ic
m
an
ag
em
en
t.
Af
ter
r
e
v
iewin
g
m
ac
h
in
e
lear
n
in
g
ap
p
r
o
ac
h
es
f
o
r
s
en
tim
en
t
an
aly
s
is
o
f
p
an
d
em
ic
twee
ts
,
we
co
n
clu
d
e
t
h
at
en
s
em
b
le
m
o
d
el
s
,
s
u
ch
as B
E
R
T
an
d
R
o
B
E
R
T
a,
ar
e
b
est f
o
r
T
witter
d
ata
[
1
4
]
.
Vac
cin
atio
n
s
ar
e
k
ey
to
f
ig
h
ti
n
g
th
e
p
a
n
d
em
ic.
So
cial
m
e
d
i
a
an
aly
tics
an
d
h
ea
lth
s
u
r
v
eillan
ce
d
ata
ar
e
u
s
ed
to
s
tu
d
y
co
m
p
licated
p
u
b
lic
p
e
r
ce
p
tio
n
s
co
n
ce
r
n
in
g
C
OVI
D
-
1
9
im
m
u
n
i
z
atio
n
s
.
Dee
p
lear
n
in
g
was
u
s
ed
to
ass
ess
m
illi
o
n
s
o
f
t
wee
ts
f
r
o
m
2
0
2
0
to
2
0
2
2
f
o
r
em
o
tio
n
s
.
T
h
is
r
esear
ch
e
x
am
in
es
th
e
p
r
eg
n
a
n
t
p
o
p
u
latio
n
'
s
em
o
tio
n
al
tr
aj
ec
t
o
r
y
an
d
t
r
en
d
s
.
W
e
id
en
tify
way
s
to
in
cr
ea
s
e
v
ac
cin
e
o
u
tr
ea
ch
b
y
co
r
r
elatin
g
s
en
tim
en
ts
with
g
l
o
b
al
v
ac
cin
atio
n
p
atter
n
s
[
1
5
]
.
So
cial
m
ed
ia
p
latf
o
r
m
s
s
aw
a
r
is
e
in
in
s
tr
u
ctiv
e
an
d
f
alse
co
n
ten
t
d
u
r
in
g
th
e
C
OVI
D
-
1
9
p
an
d
em
ic.
T
h
is
s
tu
d
y
s
tu
d
ies
T
witter
co
m
m
en
ts
o
n
C
OVI
D
-
1
9
v
ac
cin
atio
n
to
h
elp
cr
ea
te
a
v
ac
ci
n
e
ac
ce
p
tan
ce
p
o
licy
.
Af
ter
test
in
g
n
u
m
er
o
u
s
m
eth
o
d
s
,
th
e
ex
tr
a
tr
ee
class
if
ier
(
E
T
C
)
u
s
in
g
th
e
b
ag
o
f
wo
r
d
s
(
B
o
W
)
i
s
th
e
m
o
s
t
ef
f
icien
t
s
en
tim
en
t
an
aly
s
is
f
r
am
ewo
r
k
f
o
r
C
OVI
D
-
1
9
tw
ee
ts
.
Ou
r
an
aly
s
is
co
n
s
tan
tly
s
h
o
ws
a
g
r
o
win
g
e
n
d
o
r
s
em
en
t
o
f
v
ac
ci
n
atio
n
[
1
6
]
.
T
witter
s
h
o
wed
th
e
s
y
n
c
h
r
o
n
i
z
ed
p
u
b
lic
attitu
d
e
ch
an
g
es
g
en
er
at
ed
b
y
C
OVI
D
-
1
9
.
T
h
i
s
r
esear
ch
d
iv
es
in
to
Me
x
ican
b
elief
s
d
u
r
in
g
a
p
a
r
ticu
lar
ly
d
if
f
icu
lt
p
an
d
em
ic
tim
e.
T
o
m
ea
s
u
r
e
attitu
d
es,
we
tr
ain
ed
m
u
ltip
le
m
o
d
els,
two
o
f
w
h
ich
wer
e
tr
ain
e
d
in
Sp
a
n
is
h
,
u
s
in
g
a
h
y
b
r
i
d
s
em
i
-
s
u
p
er
v
is
ed
tech
n
i
q
u
e.
C
o
m
p
ar
in
g
th
e
Sp
an
is
h
-
ce
n
tr
ic
m
o
d
el
to
SVM
an
d
Dec
is
i
o
n
T
r
ee
s
s
h
o
wed
its
h
ig
h
er
ac
cu
r
ac
y
.
Me
x
ican
T
witter
u
s
er
s
'
C
OVI
D
-
1
9
m
o
o
d
s
wer
e
ass
ess
ed
u
s
i
n
g
th
is
tai
lo
r
ed
ap
p
r
o
ac
h
[
1
7
]
.
T
h
e
r
esear
ch
ex
am
i
n
ed
h
o
w
d
aily
T
ik
T
o
k
ca
s
e
f
ilm
s
b
y
P
u
b
lic
H
ea
lth
O
r
g
an
i
z
atio
n
s
(
PHAs
)
af
f
ec
ted
C
OVI
D
-
1
9
ac
ce
p
tan
ce
an
d
u
n
d
er
s
tan
d
in
g
.
T
h
e
g
o
al
was
to
u
n
d
er
s
tan
d
p
u
b
lic
o
p
in
io
n
an
d
an
x
iet
y
d
u
r
in
g
th
e
2
0
2
2
Sh
a
n
g
h
ai
lo
ck
d
o
wn
.
T
h
e
c
r
is
is
an
d
em
er
g
en
cy
r
is
k
co
m
m
u
n
icatio
n
m
o
d
el
d
i
v
id
es
th
e
lo
ck
d
o
wn
in
t
o
f
iv
e
p
a
r
ts
[
1
8
]
.
T
h
is
r
esear
ch
d
iv
id
ed
th
e
S
h
an
g
h
ai
s
h
u
t
d
o
wn
in
to
s
tag
es
.
User
r
e
ac
tio
n
to
Hea
lth
y
C
h
in
a'
s
d
aily
T
ik
T
o
k
ca
s
e
v
id
e
o
s
was
ex
ten
s
iv
e
ly
an
aly
z
ed
th
r
o
u
g
h
o
u
t
t
h
ese
p
er
io
d
s
.
T
h
e
p
r
e
-
tr
ain
ed
E
R
NI
E
m
o
d
el
class
if
ied
u
s
er
co
m
m
e
n
t
em
o
tio
n
s
.
Sen
tim
en
t
class
if
icatio
n
also
in
f
o
r
m
e
d
s
em
an
tic
n
etwo
r
k
in
v
esti
g
atio
n
s
.
Giv
e
n
th
e
h
i
g
h
co
s
t
o
f
co
n
tr
o
llin
g
th
e
ep
i
d
em
ic,
th
e
p
u
b
lic
w
as
u
n
willin
g
to
tak
e
p
r
o
p
h
y
lactic
m
ea
s
u
r
es
in
th
e
b
eg
in
n
in
g
.
Sh
an
g
h
ai'
s
u
n
ilate
r
al
d
ef
in
itio
n
o
f
"a
s
y
m
p
to
m
at
ic
p
atien
ts
"
af
f
ec
ts
co
n
tr
o
l
e
f
f
o
r
ts
in
o
th
er
m
u
n
i
cip
alities
an
d
d
aily
life
.
D
u
r
i
n
g
m
ai
n
ten
an
ce
,
in
d
iv
i
d
u
al
s
f
o
cu
s
ed
o
n
d
is
ea
s
e
-
r
elate
d
ar
ea
s
o
f
th
eir
liv
es.
I
n
ter
est
in
d
aily
ca
s
e
u
p
d
ate
v
id
eo
s
d
r
o
p
p
ed
af
ter
r
eso
lu
tio
n
.
T
h
e
ap
p
ar
en
t
d
iv
er
g
en
ce
o
f
lo
ca
l
g
o
v
er
n
m
e
n
t
s
tr
ateg
y
f
r
o
m
ce
n
tr
al
g
o
v
e
r
n
m
en
t
o
r
d
er
s
ca
u
s
ed
wid
esp
r
e
ad
d
is
p
leasu
r
e.
T
h
e
wo
r
ld
wid
e
h
ea
lth
lan
d
s
c
ap
e
is
th
r
ea
ten
ed
b
y
C
OVI
D
-
1
9
.
Vac
cin
e
d
ev
elo
p
m
e
n
t
an
d
d
is
tr
ib
u
tio
n
ar
e
cr
u
cial
to
p
r
ev
en
tin
g
th
is
.
As
v
ac
cin
e
d
i
s
s
em
in
atio
n
in
cr
ea
s
es,
v
ir
al
i
m
m
u
n
ity
s
h
o
u
ld
r
is
e.
I
n
th
e
d
ig
ital
ag
e,
T
witter
is
ess
en
tial
f
o
r
p
u
b
lic
o
p
i
n
io
n
r
esear
ch
,
esp
ec
ially
f
o
r
v
a
cc
in
a
tio
n
ef
f
o
r
ts
.
T
h
is
s
tu
d
y
an
aly
z
ed
C
OVI
D
v
ac
cin
atio
n
twee
ts
u
s
in
g
ad
v
an
ce
d
AI
an
d
g
eo
-
s
p
atial
m
eth
o
d
s
.
T
ex
tB
lo
b
a
n
aly
z
ed
th
es
e
twee
ts
'
s
en
tim
en
t
p
o
lar
ity
.
T
h
e
d
ata
was
g
r
ap
h
ed
u
s
in
g
wo
r
d
clo
u
d
s
a
n
d
e
m
o
tio
n
was
i
d
en
tifie
d
u
s
in
g
B
E
R
T
.
Geo
co
d
in
g
p
lace
d
an
d
v
is
u
ally
d
is
p
lay
ed
em
o
tio
n
d
ata
o
n
a
g
lo
b
al
m
a
p
.
Ad
v
an
ce
d
ap
p
r
o
ac
h
es
in
clu
d
e
h
o
ts
p
o
t
an
aly
s
is
an
d
k
er
n
el
d
en
s
ity
esti
m
atio
n
s
th
at
id
en
tif
y
r
eg
io
n
s
with
p
leasan
t,
n
eg
ativ
e,
o
r
n
eu
tr
al
s
en
tim
en
ts
.
T
h
e
m
o
d
el'
s
ac
cu
r
ac
y
,
r
ec
all,
an
d
F
-
s
co
r
e
f
o
r
p
o
s
itiv
e
an
d
n
e
g
a
t
iv
e
s
en
tim
en
t
ca
teg
o
r
i
z
atio
n
s
wer
e
co
m
p
ar
ed
to
well
-
estab
lis
h
ed
ap
p
r
o
ac
h
es.
T
h
e
m
ain
p
u
r
p
o
s
e
is
to
s
tu
d
y
p
u
b
lic
o
p
in
io
n
o
n
v
ac
cin
atio
n
p
r
o
g
r
am
s
d
u
r
in
g
g
lo
b
al
h
ea
lth
e
m
er
g
e
n
cies u
s
in
g
s
en
tim
en
t a
n
d
g
eo
g
r
ap
h
ical
an
aly
tics
[
1
9
]
.
T
h
e
wo
r
ld
wid
e
s
p
r
ea
d
o
f
S
AR
S
-
C
o
V
-
2
h
as
ac
ce
ler
ated
s
in
ce
Ma
r
ch
2
0
2
0
.
Millio
n
s
h
av
e
b
ee
n
in
f
ec
ted
s
in
ce
th
e
o
u
t
b
r
ea
k
,
p
r
o
m
p
tin
g
g
lo
b
al
r
ea
ctio
n
s
o
n
T
witter
.
T
h
e
m
ain
g
o
al
o
f
th
is
r
esear
ch
is
to
an
aly
z
e
Hin
d
i
-
s
p
ea
k
i
n
g
T
witter
u
s
er
s
'
v
iews
o
n
th
e
p
an
d
e
m
ic.
Data
p
r
o
ce
s
s
in
g
b
e
g
an
with
NL
P
o
n
Hin
d
i
C
OVI
D
-
1
9
twee
ts
.
Af
ter
war
d
,
a
u
n
iq
u
e
Gr
e
y
W
o
lf
o
p
tim
i
z
atio
n
ap
p
r
o
ac
h
im
p
r
o
v
ed
f
ea
tu
r
e
s
elec
tio
n
.
T
h
e
m
ajo
r
an
aly
tical
to
o
l
was
a
h
y
b
r
id
m
o
d
el
th
at
co
m
b
in
ed
t
h
e
b
en
ef
its
o
f
C
NNs
an
d
L
STM
s
.
A
co
m
p
ar
is
o
n
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
el
with
cu
r
r
en
t
m
ac
h
in
e
lear
n
in
g
p
ar
ad
i
g
m
s
s
h
o
wed
s
u
p
er
io
r
ac
cu
r
ac
y
,
p
r
ec
i
s
io
n
,
r
ec
all,
an
d
F
-
s
co
r
e
[
2
0
]
.
T
h
e
C
OVI
D
-
1
9
e
p
id
em
ic
h
as
h
ad
a
m
ajo
r
ec
o
n
o
m
ic
e
f
f
ec
t
o
n
wo
r
ld
wid
e
f
in
an
cia
l
m
ar
k
ets.
I
n
v
esto
r
s
ar
e
in
ter
ested
in
ac
cu
r
ate
f
o
r
ec
ast
s
s
in
ce
s
h
ar
es
ar
e
v
o
latile,
esp
ec
ially
d
u
r
i
n
g
ep
id
em
ics.
T
h
is
r
esear
ch
in
tr
o
d
u
ce
s
s
en
tim
en
t
an
aly
s
is
o
f
C
OVI
D
-
1
9
n
ew
s
to
p
r
e
d
ict
s
to
ck
m
a
r
k
et
c
h
an
g
es.
Daily
s
to
ck
m
o
v
em
en
ts
wer
e
f
o
r
ec
asted
u
s
in
g
C
OVI
D
-
1
9
s
to
ck
n
e
ws
h
ea
d
lin
es.
Ma
ch
in
e
lear
n
in
g
class
if
ier
s
also
esti
m
ated
th
e
ep
id
em
ic'
s
im
p
ac
t
o
n
h
ig
h
-
v
alu
e
s
to
ck
s
in
clu
d
in
g
T
esla,
I
n
c.
(
T
SLA
)
,
Am
az
o
n
.
c
o
m
,
I
n
c
(
AM
Z
N)
,
an
d
Alp
h
ab
et
I
n
c.
(
GOOG
)
.
T
o
b
o
o
s
t
f
o
r
ec
ast
a
cc
u
r
ac
y
,
f
ea
tu
r
es
wer
e
r
ef
i
n
e
d
an
d
s
p
am
twee
ts
wer
e
r
em
o
v
ed
.
T
h
e
s
y
s
tem
'
s
t
ex
tu
al
an
aly
s
is
o
f
s
o
cial
m
ed
i
a
an
d
d
ata
m
i
n
in
g
o
f
h
is
to
r
ica
l
s
to
ck
d
ata
m
ak
e
it
u
n
iq
u
e.
T
h
e
ap
p
r
o
ac
h
p
r
e
d
icted
s
to
ck
s
ac
cu
r
at
ely
[
2
1
]
.
So
cial
m
ed
ia
h
as
b
ee
n
u
tili
z
ed
g
lo
b
ally
to
s
h
ar
e
id
ea
s
,
f
ee
l
in
g
s
,
an
d
v
iewp
o
in
ts
o
n
th
e
C
OVI
D
-
1
9
p
an
d
em
ic
f
r
o
m
its
s
tar
t.
T
witt
er
an
d
o
th
er
d
i
g
ital
p
latf
o
r
m
s
s
av
e
p
u
b
licly
co
m
m
u
n
icate
d
m
ater
ial,
allo
win
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
8
1
4
I
n
t J Ad
v
Ap
p
l Sci
,
Vo
l.
14
,
No
.
2
,
J
u
n
e
2
0
2
5
:
5
6
2
-
579
566
in
d
iv
id
u
als
to
d
eb
ate
th
e
p
an
d
em
ic
at
an
y
tim
e
o
r
lo
ca
tio
n
.
T
h
e
f
ast
r
is
e
in
ca
s
es
wo
r
ld
wid
e
h
as
ca
u
s
ed
p
eo
p
le
t
o
wo
r
r
y
,
d
r
ea
d
,
an
d
f
ee
l
u
n
c
o
m
f
o
r
tab
le.
T
h
is
s
tu
d
y
p
r
o
p
o
s
es
a
n
o
v
el
s
en
tim
en
t
a
n
aly
s
is
m
eth
o
d
f
o
r
r
ec
o
g
n
i
z
in
g
em
o
tio
n
s
in
Mo
r
o
cc
an
C
OVI
D
-
1
9
twee
ts
f
r
o
m
Ma
r
ch
to
Octo
b
er
2
0
2
0
.
O
u
r
s
y
s
tem
class
if
ie
s
twee
ts
a
s
p
o
s
itiv
e,
b
ad
,
o
r
n
eu
tr
al
u
s
in
g
r
ec
o
m
m
en
d
ati
o
n
s
.
Ou
r
m
eth
o
d
o
u
tp
er
f
o
r
m
s
well
-
estab
li
s
h
ed
m
ac
h
in
e
lear
n
in
g
m
eth
o
d
s
with
8
6
%
ac
cu
r
ac
y
,
ac
co
r
d
in
g
t
o
ex
p
er
im
e
n
ts
.
T
em
p
o
r
al
ch
a
n
g
es
in
s
en
tim
en
ts
s
h
o
w
th
at
th
e
s
h
if
tin
g
C
OVI
D
-
1
9
s
ce
n
ar
io
in
th
e
c
o
u
n
tr
y
af
f
ec
ted
p
u
b
lic
f
ee
li
n
g
s
[
2
2
]
.
T
witter
d
ata
f
r
o
m
Ma
r
ch
2
0
2
3
was u
s
ed
to
an
aly
z
e
I
n
d
ian
p
u
b
lic
o
p
in
io
n
o
n
th
e
C
OVI
D
-
1
9
v
ac
cin
e.
B
ef
o
r
e
s
en
tim
en
t
an
aly
s
is
u
s
in
g
NL
P
,
twee
ts
wer
e
p
r
ep
r
o
c
ess
ed
an
d
s
an
iti
z
ed
u
s
in
g
r
el
ev
an
t
h
ash
tag
s
an
d
k
ey
wo
r
d
s
.
N
u
m
er
o
u
s
twee
ts
s
u
p
p
o
r
tin
g
im
m
u
n
i
z
atio
n
s
h
o
w
a
p
o
s
itiv
e
o
u
tlo
o
k
.
H
o
wev
er
,
q
u
esti
o
n
s
wer
e
r
aised
co
n
ce
r
n
in
g
i
m
m
u
n
i
z
ati
o
n
r
elu
cta
n
ce
,
s
id
e
e
f
f
ec
ts
,
a
n
d
g
o
v
er
n
m
en
t
a
n
d
p
h
ar
m
ac
eu
tical
m
is
tr
u
s
t.
A
m
o
r
e
d
etailed
an
aly
s
is
o
f
v
ie
ws
b
y
g
en
d
er
,
ag
e
,
an
d
r
e
g
io
n
al
in
d
icato
r
s
r
ev
ea
led
d
if
f
er
en
t
f
ee
lin
g
s
ac
r
o
s
s
p
o
p
u
latio
n
g
r
o
u
p
s
.
Ou
r
a
n
aly
s
is
illu
m
in
ates
th
e
v
ac
cin
atio
n
s
itu
atio
n
in
I
n
d
ia
an
d
em
p
h
asi
z
es
th
e
n
ec
ess
ity
f
o
r
p
er
s
o
n
ali
z
ed
c
o
m
m
u
n
icatio
n
to
m
in
im
i
z
e
v
ac
cin
e
u
n
c
er
tain
ty
an
d
b
o
o
s
t
ac
ce
p
tab
il
ity
am
o
n
g
ch
o
s
en
p
o
p
u
latio
n
s
[
2
3
]
.
US
C
OVI
D
-
1
9
s
tati
s
tics
.
Af
r
ican
Am
er
ican
s
h
ad
a
d
is
p
r
o
p
o
r
tio
n
ate
s
h
ar
e
o
f
v
ir
al
in
f
e
ctio
n
s
an
d
d
ea
th
s
,
ac
co
r
d
in
g
t
o
ch
a
n
g
e
d
ata
ca
p
tu
r
e
(
C
DC
)
d
ata
th
r
o
u
g
h
J
u
n
e
2
0
2
0
.
T
h
is
m
is
m
atch
h
ig
h
lig
h
ts
th
e
n
ee
d
to
u
n
d
er
s
tan
d
Af
r
ican
Am
er
ic
an
C
OVI
D
-
1
9
ex
p
er
ien
ce
s
an
d
v
iews.
Ou
r
s
tu
d
y
ex
am
in
es
Af
r
ican
Am
er
ican
p
an
d
em
ic
n
ar
r
ativ
es
u
s
in
g
a
s
p
ec
t
-
b
ased
s
en
tim
en
t
a
n
aly
s
is
o
f
2
0
2
0
T
witter
d
ata.
O
u
r
m
ac
h
i
n
e
lear
n
in
g
tech
n
o
lo
g
y
f
ilter
s
twee
ts
th
at
ar
e
u
n
r
elate
d
to
C
OVI
D
-
1
9
o
r
n
o
t
f
r
o
m
Af
r
ica
n
Am
er
ica
n
s
.
Ap
p
r
o
x
im
atel
y
4
m
illi
o
n
twee
ts
wer
e
an
aly
z
ed
.
T
h
e
twee
ts
wer
e
m
o
s
tly
g
lo
o
m
y
,
with
v
o
lu
m
e
s
p
ik
es
co
in
cid
in
g
with
US
p
an
d
em
ics.
T
h
is
r
esear
ch
s
h
o
ws
h
o
w
p
an
d
em
ic
lan
g
u
ag
e
ch
an
g
ed
d
u
r
in
g
th
e
y
ea
r
an
d
h
ig
h
lig
h
ts
cr
u
cial
ch
allen
g
es
in
clu
d
in
g
f
o
o
d
s
h
o
r
tag
es
an
d
v
ac
ci
n
e
r
ejec
tio
n
.
T
o
u
n
d
er
s
tan
d
h
o
w
th
e
p
a
n
d
e
m
ic
af
f
ec
ted
Af
r
ica
n
Am
er
ican
T
witter
u
s
er
s
'
co
n
v
er
s
atio
n
s
,
we
wis
h
to
s
tr
ess
wo
r
d
li
n
k
ag
es
a
n
d
attitu
d
es
[
2
4
]
.
T
h
e
m
e
d
ia
m
u
s
t
ag
g
r
ess
iv
ely
p
r
o
m
o
te
h
ea
lth
c
o
n
ce
r
n
s
to
r
aise p
u
b
lic
awa
r
e
n
ess
an
d
r
e
d
u
ce
h
ea
lth
h
az
ar
d
s
to
im
p
r
o
v
e
s
o
ciety
.
Dee
p
n
eu
r
al
n
etwo
r
k
s
ar
e
b
e
co
m
in
g
in
c
r
ea
s
in
g
ly
p
o
p
u
lar
in
tex
tu
al
s
en
tim
en
t
an
aly
s
is
,
allo
win
g
r
ea
l
-
tim
e
h
ea
lth
m
o
n
ito
r
in
g
an
d
in
s
ig
h
t
s
.
C
o
v
-
Att
-
B
iLST
M
i
s
an
in
n
o
v
ativ
e
ar
tific
ial
in
tellig
en
ce
m
o
d
el
f
o
r
s
en
tim
en
t
an
a
ly
s
is
o
f
C
OVI
D
-
1
9
n
ews
h
ea
d
lin
es.
T
h
e
m
o
d
el
u
s
es
d
ee
p
n
eu
r
al
n
etwo
r
k
s
.
Ou
r
wo
r
k
u
s
es
atten
tio
n
p
r
o
ce
s
s
es,
em
b
ed
d
in
g
m
eth
o
d
s
,
an
d
s
em
an
tic
d
ata
lab
eli
n
g
to
im
p
r
o
v
e
p
r
e
d
ictio
n
ac
cu
r
ac
y
.
Ou
r
m
o
d
el
p
er
f
o
r
m
ed
well
v
er
s
u
s
o
th
er
m
ac
h
in
e
lear
n
in
g
class
if
ier
s
,
atta
in
in
g
0
.
9
3
1
test
ac
cu
r
ac
y
.
W
e
also
ap
p
lied
it
to
7
3
,
1
3
8
p
a
n
d
em
ic
twee
ts
f
r
o
m
s
ix
f
o
r
eig
n
s
o
u
r
ce
s
to
a
cc
u
r
ately
d
e
p
ict
g
lo
b
al
C
O
VI
D
-
1
9
n
ews
an
d
v
ac
cin
atio
n
d
is
cu
s
s
io
n
s
[
2
5
]
.
Pro
g
r
ess
is
d
r
iv
en
b
y
k
n
o
wled
g
e.
Data
co
llectio
n
tec
h
n
iq
u
es
h
a
v
e
ev
o
lv
ed
la
r
g
e
ly
d
u
e
t
o
tech
n
o
lo
g
ical
a
d
v
an
ce
m
en
t.
Ma
n
y
tech
n
o
lo
g
ical
m
e
d
iu
m
s
ar
e
f
ast,
r
eliab
le,
a
n
d
e
f
f
ic
ien
t,
y
et
o
t
h
er
s
ar
e
lack
in
g
f
o
r
v
ar
io
u
s
r
ea
s
o
n
s
.
T
h
is
r
esear
ch
will
an
aly
z
e
Nig
er
ian
C
OVI
D
-
1
9
twee
ts
an
d
p
u
b
lic
o
p
in
io
n
u
s
in
g
T
ex
tB
lo
b
an
d
VADE
R
m
o
d
els.
I
t
m
ea
s
u
r
es
e
m
o
tio
n
al
r
esp
o
n
s
es
an
d
illu
m
in
ates
Nig
er
ia's
s
o
cieta
l
,
ec
o
lo
g
ical,
an
d
ec
o
n
o
m
ic
ef
f
ec
ts
.
T
h
is
wo
r
k
m
i
g
h
t
b
e
u
s
e
f
u
l
f
o
r
d
ata
s
cien
ce
,
m
ac
h
i
n
e
lear
n
in
g
,
an
d
d
ee
p
lear
n
in
g
r
esear
ch
er
s
.
Af
ter
p
r
e
p
r
o
ce
s
s
in
g
1
,
0
4
8
,
5
7
5
'
C
OVI
D
-
1
9
'
twee
ts
,
T
ex
tB
lo
b
an
d
VA
DE
R
wer
e
u
s
ed
t
o
ass
es
s
s
en
tim
en
t.
Ou
r
r
esear
ch
f
o
u
n
d
a
v
ar
iety
o
f
p
er
s
p
ec
t
iv
es,
d
em
o
n
s
tr
atin
g
th
e
ef
f
ica
cy
o
f
s
o
cial
m
e
d
ia
an
aly
s
is
in
ass
is
tin
g
m
ajo
r
co
m
p
an
ies b
attle
C
OVI
D
-
1
9
'
s
e
f
f
ec
ts
an
d
m
is
in
f
o
r
m
atio
n
[
2
6
]
.
T
witter
allo
ws
p
eo
p
le
g
lo
b
ally
to
co
m
m
u
n
icate
t
h
eir
id
ea
s
,
esp
ec
ially
d
u
r
in
g
m
ajo
r
ev
e
n
ts
lik
e
th
e
cu
r
r
en
t
p
a
n
d
em
ic.
T
h
is
r
ese
ar
ch
u
s
es
twee
ts
to
an
aly
z
e
I
n
d
ian
p
er
ce
p
tio
n
s
co
n
ce
r
n
i
n
g
C
OVI
D
-
1
9
an
d
im
m
u
n
i
z
atio
n
.
W
e
class
if
ied
twee
t
em
o
tio
n
s
u
s
in
g
d
ee
p
lea
r
n
in
g
a
n
d
lex
ic
o
n
-
ce
n
ter
ed
m
eth
o
d
s
.
T
h
e
le
x
ico
n
ap
p
r
o
ac
h
u
tili
z
ed
VADE
R
an
d
NR
C
L
ex
,
wh
er
ea
s
th
e
d
ee
p
lear
n
in
g
m
et
h
o
d
u
s
ed
B
i
-
L
STM
an
d
g
ated
r
ec
u
r
r
en
t
u
n
it
(
GR
U
)
,
y
ield
i
n
g
am
az
in
g
ac
c
u
r
ac
y
.
Ou
r
m
o
d
els
m
ay
h
elp
h
ea
lth
ca
r
e
p
r
o
f
ess
io
n
als
an
d
d
ec
is
io
n
-
m
ak
er
s
in
f
u
tu
r
e
p
an
d
em
ics
[2
7]
.
So
cial
d
is
tan
cin
g
m
ea
s
u
r
es
d
i
d
n
o
t
d
ete
r
2
0
%
o
f
co
m
m
u
ter
s
f
r
o
m
u
s
in
g
p
u
b
lic
tr
a
n
s
p
o
r
t
d
u
r
in
g
th
e
C
OVI
D
-
1
9
ep
id
em
ic.
T
r
a
d
itio
n
al
u
r
b
an
tr
a
n
s
p
o
r
tatio
n
d
at
a
co
llectio
n
tech
n
iq
u
es
m
is
s
p
ass
en
g
er
s
'
co
m
p
lex
p
s
y
ch
o
lo
g
ical
ex
p
e
r
ien
ce
s
.
T
h
u
s
,
u
n
d
er
s
tan
d
in
g
t
h
e
tr
an
s
p
o
r
tatio
n
-
d
ep
en
d
en
t
im
p
o
v
e
r
is
h
ed
p
o
p
u
latio
n
s
'
s
itu
atio
n
is
to
u
g
h
.
W
e
u
s
ed
m
ac
h
in
e
lear
n
i
n
g
t
o
s
eg
m
e
n
t
T
witter
d
ata
to
p
r
o
p
er
ly
d
ep
ict
t
h
e
tr
av
el
p
atter
n
s
o
f
1
2
0
,
0
0
0
m
et
r
o
Van
co
u
v
er
t
r
an
s
it
r
id
er
s
b
ef
o
r
e
an
d
th
r
o
u
g
h
o
u
t
th
e
p
an
d
em
ic.
O
u
r
r
esear
ch
f
o
u
n
d
a
co
n
s
id
er
ab
le
r
is
e
in
u
n
f
a
v
o
r
a
b
le
v
iews,
n
o
tab
ly
ac
r
o
s
s
d
e
m
o
g
r
ap
h
ic
ca
teg
o
r
ies,
d
u
r
in
g
th
e
ea
r
ly
C
OVI
D
-
1
9
ep
id
em
ic.
Ou
r
g
o
al
is
to
id
en
t
if
y
tr
an
s
it
u
s
er
s
'
in
eq
u
ities
an
d
r
is
k
s
d
u
r
i
n
g
th
e
cr
is
is
to
im
p
r
o
v
e
p
u
b
lic
h
ea
lth
an
d
tr
an
s
p
o
r
t p
lan
n
in
g
in
f
u
tu
r
e
d
is
r
u
p
tio
n
s
[
2
8
]
.
T
h
is
s
tu
d
y
u
s
es
NL
P
an
d
o
p
in
io
n
a
n
aly
s
is
(
SA)
to
ass
es
s
I
talian
p
u
b
lic
o
p
in
i
o
n
ab
o
u
t
C
OVI
D
-
19
im
m
u
n
i
z
atio
n
.
T
h
is
r
esear
ch
s
o
lely
in
clu
d
es
I
talian
v
ac
cin
e
twee
ts
f
r
o
m
J
an
u
ar
y
2
0
2
1
to
Feb
r
u
ar
y
2
0
2
2
.
Fro
m
1
,
6
0
2
,
9
4
0
twee
ts
in
clu
d
in
g
"v
ac
cin
e
,
"
3
5
3
,
2
1
7
wer
e
ex
am
in
ed
.
W
e
d
iv
i
d
e
u
s
er
s
in
to
f
o
u
r
ca
teg
o
r
ies:
co
m
m
o
n
u
s
er
s
,
m
ed
ia,
m
ed
ic
in
e,
an
d
p
o
liti
cs
,
wh
ich
is
n
o
v
el.
T
h
is
ca
teg
o
r
i
z
atio
n
e
v
a
lu
ates
u
s
er
p
r
o
f
iles
u
s
in
g
NL
P
an
d
d
o
m
ain
-
s
p
ec
if
ic
lex
ico
n
s
.
Ou
r
s
en
tim
en
t
an
aly
s
is
u
s
e
s
I
talian
'
s
p
o
lar
ized
a
n
d
h
ea
ted
wo
r
d
s
to
r
ev
ea
l
ea
ch
u
s
er
g
r
o
u
p
'
s
to
n
e
.
T
h
e
s
tu
d
y
f
o
u
n
d
g
e
n
er
ally
n
eg
ativ
e
attitu
d
es
th
r
o
u
g
h
o
u
t
th
e
r
ev
iew
p
er
io
d
,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ad
v
Ap
p
l Sci
I
SS
N:
2252
-
8
8
1
4
S
en
timen
t a
n
a
lysi
s
o
f v
a
cc
in
e
d
a
ta
u
s
in
g
en
h
a
n
ce
d
d
ee
p
lea
r
n
in
g
a
l
g
o
r
ith
ms
(
Mo
n
ika
V
erma
)
567
esp
ec
ially
am
o
n
g
co
m
m
o
n
u
s
er
s
.
I
n
p
ar
ticu
lar
,
p
o
s
t
-
im
m
u
n
izatio
n
f
atalities
alter
ed
att
itu
d
e
p
atter
n
s
af
ter
1
4
m
o
n
th
s
[
2
9
]
.
Sen
tim
en
t
an
aly
s
is
in
NL
P
ma
y
ex
tr
ac
t
u
s
ef
u
l
in
f
o
r
m
atio
n
f
r
o
m
o
n
lin
e
C
OVI
D
-
1
9
d
ata,
h
elp
in
g
C
h
in
a
f
ig
h
t
th
e
p
a
n
d
em
ic.
D
ee
p
lear
n
in
g
-
b
ased
s
en
tim
en
t
an
aly
s
is
alg
o
r
ith
m
s
h
av
e
i
m
p
r
o
v
e
d
;
h
o
wev
er
,
d
ataset
lim
its
ty
p
ically
h
in
d
er
th
em
.
W
e
p
r
esen
t
a
f
ed
er
ated
lear
n
in
g
s
y
s
tem
th
at
b
len
d
s
B
E
R
T
with
a
m
u
lti
-
s
ca
le
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
e
two
r
k
(
Fed
_
B
E
R
T
_
MSC
NN)
to
ad
d
r
ess
th
is
p
r
o
b
lem
.
T
h
e
s
u
g
g
ested
m
o
d
el
u
s
es
tr
an
s
f
o
r
m
er
-
d
er
iv
e
d
b
id
ir
ec
tio
n
al
en
co
d
er
r
e
p
r
esen
ta
tio
n
s
an
d
m
u
ltis
ca
le
co
n
v
o
lu
tio
n
lay
er
s
.
T
h
e
f
ed
er
ated
ar
ch
itectu
r
e
tr
ain
s
d
atasets
in
d
ep
en
d
en
tly
u
s
in
g
a
ce
n
tr
al
s
er
v
er
an
d
lo
ca
l
d
ee
p
lear
n
in
g
s
y
s
tem
s
.
E
d
g
e
n
etwo
r
k
s
s
im
p
lify
m
o
d
el
in
teg
r
atio
n
an
d
p
a
r
am
ete
r
ex
ch
an
g
e
.
T
h
is
n
ew
n
etwo
r
k
ad
d
r
ess
es
d
ata
s
ca
r
city
,
in
cr
ea
s
es
co
m
m
u
n
icatio
n
ef
f
icien
c
y
,
an
d
i
m
p
r
o
v
es
d
ata
p
r
iv
ac
y
d
u
r
in
g
tr
ain
in
g
.
T
h
e
Fed
_
B
E
R
T
_
MSC
NN
m
o
d
el
o
u
tp
er
f
o
r
m
ed
its
p
ee
r
s
o
n
s
ix
s
o
cial
s
ites
[
3
0
]
.
T
h
e
ex
tr
ao
r
d
in
ar
y
C
OVI
D
-
1
9
p
an
d
em
ic
a
f
f
ec
ted
f
i
n
an
cial
m
ar
k
ets,
esp
ec
ially
in
its
ea
r
ly
s
tag
es.
T
h
e
cu
r
r
en
t
r
esear
ch
ex
a
m
in
es
h
o
w
C
OVI
D
-
1
9
n
ews
f
l
o
ws
af
f
ec
t
m
ar
k
et
ex
p
ec
tatio
n
s
.
W
e
ex
am
in
ed
2
0
3
,
8
8
6
C
OVI
D
-
19
r
elate
d
in
ter
n
et
ar
ticles
f
r
o
m
J
a
n
u
ar
y
to
J
u
n
e
2
0
2
0
o
n
Ma
r
k
etW
atch
.
co
m
,
NYT
im
es.c
o
m
,
an
d
R
eu
ter
s
.
co
m
.
W
e
u
s
ed
a
f
in
an
cially
-
tu
n
ed
B
E
R
T
m
o
d
el,
wh
ich
u
n
d
e
r
s
tan
d
s
co
n
tex
tu
al
wo
r
d
s
em
an
tics
,
t
o
a
n
aly
z
e
th
ese
ar
ticles'
s
en
tim
en
t
s
u
s
in
g
m
ac
h
in
e
lear
n
i
n
g
.
Ou
r
r
esear
ch
s
h
o
ws
a
s
u
b
s
tan
tial
p
o
s
itiv
e
co
r
r
elatio
n
b
etwe
en
s
en
tim
en
t
in
d
ic
ato
r
s
an
d
S&
P
5
0
0
m
ar
k
et
p
er
f
o
r
m
a
n
ce
.
I
t'
s
in
ter
esti
n
g
th
at
NYT
im
es.c
o
m
'
s
at
titu
d
e
an
d
n
ews k
in
d
s
af
f
e
cted
m
ar
k
et
p
e
r
f
o
r
m
an
ce
d
if
f
e
r
en
tly
[
3
1
]
.
3.
P
RO
P
O
SE
D
M
E
T
H
O
DO
L
O
G
Y
3
.
1
.
P
r
o
po
s
ed
f
lo
wcha
rt
F
ig
u
r
e
1
p
r
esen
ts
a
co
m
p
r
e
h
e
n
s
iv
e
wo
r
k
f
lo
w
f
o
r
an
aly
zin
g
tex
tu
al
d
ata
u
s
in
g
an
ad
v
a
n
ce
d
B
i
-
L
STM
m
o
d
el.
T
h
e
p
r
o
ce
s
s
b
eg
in
s
wi
th
a
r
aw
te
x
t
in
p
u
t
th
at
is
f
ir
s
t
s
u
b
jecte
d
to
p
r
o
ce
s
s
in
g
an
d
c
lean
in
g
to
r
em
o
v
e
n
o
is
e,
an
d
ir
r
elev
an
t
in
f
o
r
m
at
io
n
,
an
d
s
tan
d
ar
d
ize
th
e
tex
t
f
o
r
an
aly
s
is
.
Nex
t,
th
e
clea
n
ed
tex
t
u
n
d
er
g
o
es
a
co
m
p
u
tatio
n
al
p
h
ase
wh
er
e
s
en
tim
en
t
s
co
r
es
ar
e
ca
lcu
late
d
,
an
d
th
e
o
b
jectiv
ity
o
f
th
e
tex
t
is
ass
es
s
ed
to
d
eter
m
in
e
th
e
s
u
b
jectiv
e
o
r
o
b
jectiv
e
n
atu
r
e
o
f
th
e
co
n
ten
t.
T
h
ese
s
en
tim
en
t
an
d
o
b
jectiv
ity
s
co
r
es
ar
e
th
en
p
r
io
r
itized
an
d
r
a
n
k
ed
in
o
r
d
er
o
f
th
ei
r
s
ig
n
if
ican
ce
t
o
r
ef
in
e
th
e
an
aly
s
is
.
Fo
llo
win
g
t
h
is
,
th
e
s
co
r
es
ar
e
u
tili
ze
d
to
i
n
itialize
a
B
i
-
L
STM
m
o
d
el,
wh
ich
in
c
o
r
p
o
r
at
es
a
co
n
te
x
tu
al
p
r
ed
ictio
n
m
e
th
o
d
,
en
h
a
n
cin
g
th
e
m
o
d
el’
s
ab
ilit
y
to
u
n
d
er
s
tan
d
co
n
tex
t
a
n
d
d
ep
e
n
d
en
cies
i
n
t
h
e
tex
t
d
ata.
T
h
is
in
itialized
B
i
-
L
STM
m
o
d
el
is
th
en
em
p
l
o
y
ed
to
ac
c
u
r
ately
id
en
tify
th
e
r
ev
iew
ca
teg
o
r
y
o
f
th
e
tex
t,
ca
teg
o
r
izin
g
it
b
a
s
ed
o
n
p
r
ed
e
f
in
e
d
cr
iter
ia
o
r
th
em
es.
T
h
e
en
tire
p
r
o
ce
s
s
cu
lm
in
ates
in
p
r
o
d
u
ci
n
g
a
f
in
al
r
esu
l
t,
wh
ich
ef
f
ec
tiv
ely
ca
teg
o
r
izes
th
e
tex
t a
n
d
p
r
o
v
id
es m
ea
n
in
g
f
u
l i
n
s
ig
h
ts
d
er
iv
ed
f
r
o
m
t
h
e
s
en
tim
en
t a
n
d
c
o
n
tex
t
o
f
th
e
in
p
u
t
d
ata.
Fig
u
r
e
1
.
Pro
p
o
s
ed
m
et
h
o
d
o
lo
g
y
wo
r
k
f
lo
w
3
.
2
.
P
r
o
po
s
ed
a
lg
o
rit
hm
T
h
e
p
r
ep
r
o
ce
s
s
in
g
alg
o
r
ith
m
tak
es
tex
tu
al
d
ata
as
in
p
u
t
an
d
p
r
o
d
u
ce
s
p
r
ep
r
o
ce
s
s
ed
tex
t
as
o
u
tp
u
t.
As
s
h
o
wn
in
A
lg
o
r
ith
m
1
.
I
t
b
eg
in
s
b
y
iter
atin
g
th
r
o
u
g
h
e
ac
h
lin
e
o
f
tex
t
in
th
e
f
ile,
w
h
er
e
th
e
tex
t
is
f
ir
s
t
to
k
en
ized
to
b
r
ea
k
it
d
o
wn
in
to
in
d
iv
i
d
u
al
w
o
r
d
s
o
r
to
k
en
s
.
Nex
t,
co
m
m
o
n
s
to
p
p
in
g
wo
r
d
s
ar
e
r
em
o
v
ed
to
elim
in
ate
u
n
n
ec
ess
ar
y
o
r
i
r
r
el
ev
an
t
wo
r
d
s
th
at
d
o
n
o
t
co
n
tr
ib
u
te
s
ig
n
if
ican
tly
to
th
e
an
al
y
s
is
.
T
h
e
tex
t
th
e
n
u
n
d
er
g
o
es
s
tem
m
in
g
,
wh
ic
h
r
ed
u
ce
s
wo
r
d
s
to
t
h
eir
r
o
o
t
f
o
r
m
,
f
o
llo
wed
b
y
lem
m
atiza
t
io
n
,
wh
ich
f
u
r
th
e
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
8
1
4
I
n
t J Ad
v
Ap
p
l Sci
,
Vo
l.
14
,
No
.
2
,
J
u
n
e
2
0
2
5
:
5
6
2
-
579
568
r
ef
in
es
th
e
tex
t
b
y
co
n
v
er
tin
g
wo
r
d
s
to
th
eir
b
ase
o
r
d
ictio
n
ar
y
f
o
r
m
.
T
h
e
p
r
ep
r
o
ce
s
s
ed
tex
t
is
th
en
r
etu
r
n
ed
f
o
r
f
u
r
th
er
a
n
aly
s
is
.
T
h
e
p
r
o
p
o
s
ed
B
i
-
L
STM
m
o
d
el
alg
o
r
ith
m
tak
es
p
r
ep
r
o
ce
s
s
ed
tex
t
as
in
p
u
t
an
d
o
u
tp
u
ts
a
class
lab
el
b
y
s
y
s
tem
atica
lly
p
r
o
ce
s
s
in
g
th
e
tex
t th
r
o
u
g
h
m
u
ltip
le
s
tep
s
to
en
h
an
ce
its
u
n
d
er
s
tan
d
in
g
a
n
d
class
if
icatio
n
a
s
s
h
o
wn
in
Alg
o
r
ith
m
2
.
I
t
b
eg
in
s
b
y
iter
atin
g
o
v
er
ea
ch
t
o
k
en
in
t
h
e
tex
t,
wh
ich
r
ep
r
ese
n
ts
a
lis
t
o
f
f
ea
tu
r
e
m
ap
s
.
First,
th
e
tex
t
is
em
b
e
d
d
ed
u
s
in
g
a
B
i
-
L
STM
em
b
ed
d
in
g
la
y
er
to
ca
p
tu
r
e
c
o
n
tex
tu
al
in
f
o
r
m
atio
n
.
T
h
e
em
b
ed
d
e
d
tex
t
is
th
en
p
r
o
ce
s
s
ed
th
r
o
u
g
h
a
B
i
-
L
STM
lay
er
to
u
n
d
e
r
s
tan
d
th
e
s
eq
u
en
ce
an
d
ca
p
tu
r
e
lo
n
g
-
r
an
g
e
d
ep
en
d
en
cies
b
etwe
en
wo
r
d
s
.
An
atten
tio
n
m
ec
h
an
is
m
is
ap
p
lied
to
f
o
cu
s
o
n
th
e
m
o
s
t
r
elev
an
t
p
ar
ts
o
f
th
e
tex
t,
im
p
r
o
v
in
g
th
e
m
o
d
el's
ab
ilit
y
to
id
e
n
tify
cr
u
cial
f
ea
tu
r
es.
Sen
tim
en
t
an
aly
s
i
s
is
co
n
d
u
cted
to
d
eter
m
in
e
th
e
p
o
lar
ity
an
d
o
b
jectiv
ity
(
Pl,
Ol)
o
f
th
e
tex
t,
h
ig
h
lig
h
tin
g
s
ig
n
if
ican
t
s
en
tim
en
t
f
ea
tu
r
es
b
y
r
an
k
in
g
th
ese
s
co
r
es.
T
h
e
co
n
tex
t
o
f
t
h
e
tex
t
is
e
x
tr
ac
ted
u
s
in
g
co
n
tex
t
ex
tr
ac
tio
n
tech
n
i
q
u
es
to
u
n
d
er
s
tan
d
th
e
u
n
d
er
ly
in
g
s
ce
n
a
r
io
a
n
d
to
p
ic.
Fin
ally
,
a
B
i
-
L
STM
-
b
ased
class
if
ier
u
s
es
th
e
s
en
tim
en
t
an
d
co
n
tex
t
f
ea
tu
r
es
(
Pl,
Ol,
an
d
C
t)
to
m
ak
e
a
f
in
al
class
if
icatio
n
,
an
d
th
e
co
r
r
esp
o
n
d
i
n
g
class
lab
el
is
r
etu
r
n
ed
as
th
e
o
u
tp
u
t.
T
h
is
ap
p
r
o
ac
h
l
ev
er
a
g
es
th
e
co
m
b
in
ed
s
tr
en
g
th
o
f
B
i
-
L
STM
an
d
atten
tio
n
m
ec
h
an
is
m
s
f
o
r
ef
f
ec
tiv
e
tex
t c
lass
if
icatio
n
b
ased
o
n
b
o
th
s
en
tim
en
t a
n
d
c
o
n
tex
tu
al
i
n
f
o
r
m
atio
n
.
T
h
e
p
r
o
p
o
s
ed
f
in
al
B
i
-
L
STM
in
-
d
ep
th
m
o
d
el
is
d
esig
n
ed
to
class
if
y
s
en
tim
en
t
in
tex
t
d
ata
ab
o
u
t
v
ac
cin
at
io
n
b
y
lev
er
a
g
in
g
m
u
ltip
le
ad
v
a
n
ce
d
n
atu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
tech
n
i
q
u
es
as
s
h
o
wn
in
Alg
o
r
ith
m
3
.
I
t
b
eg
in
s
with
a
p
r
ep
r
o
ce
s
s
in
g
s
tep
wh
er
e
r
a
w
tex
t
d
ata
is
clea
n
ed
b
y
r
em
o
v
in
g
n
o
is
e
s
u
ch
as
UR
L
s
an
d
s
p
ec
ial
ch
ar
ac
ter
s
an
d
n
o
r
m
alize
d
b
y
co
n
v
er
tin
g
tex
t
to
lo
wer
ca
s
e
an
d
a
p
p
ly
i
n
g
s
tem
m
in
g
.
Nex
t,
th
e
p
r
e
p
r
o
ce
s
s
ed
te
x
t
is
tr
a
n
s
f
o
r
m
ed
in
to
a
n
u
m
er
ical
f
o
r
m
a
t
u
s
in
g
wo
r
d
em
b
e
d
d
in
g
s
lik
e
Glo
Ve,
c
o
n
v
e
r
tin
g
ea
ch
to
k
e
n
in
to
a
d
en
s
e
v
ec
to
r
r
ep
r
esen
tatio
n
.
T
h
e
s
eq
u
e
n
ce
o
f
wo
r
d
v
ec
t
o
r
s
is
th
en
p
ass
ed
th
r
o
u
g
h
a
Bi
-
L
ST
M
lay
er
,
wh
ich
ca
p
tu
r
es
b
o
th
f
o
r
wa
r
d
an
d
b
ac
k
w
ar
d
co
n
tex
t
u
al
d
ep
en
d
en
cies
to
u
n
d
er
s
tan
d
th
e
s
eq
u
en
ce
an
d
m
ea
n
in
g
with
in
th
e
tex
t.
An
atten
tio
n
m
ec
h
a
n
is
m
is
ap
p
lied
to
th
e
h
id
d
e
n
s
tates
g
en
er
ated
b
y
th
e
B
i
-
L
STM
,
h
ig
h
lig
h
tin
g
th
e
m
o
s
t
im
p
o
r
tan
t
p
ar
ts
o
f
th
e
tex
t
f
o
r
an
aly
s
is
.
Sen
tim
en
t
an
aly
s
is
is
p
er
f
o
r
m
ed
o
n
th
ese
tex
t
f
ea
tu
r
es
to
p
r
ed
ict
s
en
tim
en
t
p
o
lar
ity
an
d
o
b
jectiv
ity
s
co
r
es,
d
eter
m
in
in
g
wh
eth
er
th
e
tex
t
ex
p
r
ess
es
p
o
s
itiv
e
o
r
n
e
g
ativ
e
s
en
tim
en
t
an
d
its
lev
el
o
f
o
b
jectiv
ity
.
Simu
ltan
e
o
u
s
ly
,
co
n
tex
t
ex
t
r
ac
tio
n
tech
n
iq
u
es
ar
e
a
p
p
lied
t
o
u
n
d
e
r
s
tan
d
th
e
to
p
ic
an
d
th
em
atic
elem
en
ts
o
f
th
e
tex
t
.
Fin
ally
,
t
h
e
s
en
tim
en
t
s
co
r
es
an
d
co
n
tex
t
f
ea
tu
r
es
ar
e
f
ed
i
n
to
a
B
i
-
L
STM
-
b
ased
class
if
ier
,
wh
ich
o
u
t
p
u
ts
a
s
en
tim
en
t
lab
el,
ca
teg
o
r
izin
g
th
e
tex
t
as
p
o
s
itiv
e
o
r
n
eg
at
iv
e
r
eg
ar
d
in
g
v
ac
cin
atio
n
.
T
h
is
co
m
p
r
eh
e
n
s
iv
e
ap
p
r
o
ac
h
co
m
b
in
es
p
r
ep
r
o
ce
s
s
in
g
,
em
b
ed
d
in
g
,
d
ee
p
lear
n
in
g
,
a
tten
tio
n
m
ec
h
an
is
m
s
,
s
en
tim
en
t
an
aly
s
is
,
an
d
co
n
tex
t
ex
tr
ac
tio
n
to
ac
h
iev
e
p
r
ec
is
e
s
en
tim
en
t c
lass
if
icatio
n
.
Alg
o
r
ith
m
1
.
Pre
p
r
o
ce
s
s
in
g
I
n
p
u
t: T
e
x
tu
a
l d
ata
Ou
tp
u
t: Pr
ep
r
o
ce
s
s
ed
tex
t
Step
1
: Beg
in
(
)
{
Step
2
:
Fo
r
ea
ch
tex
t lin
e
in
th
e
f
ile:
Step
3
: T
ex
t
←
E
x
tr
ac
t to
k
en
s
Step
4
:
T
ex
t
←
R
em
o
v
e
s
to
p
p
i
n
g
wo
r
d
s
Step
5
: T
ex
t
←
Stem
m
in
g
Step
6
: T
ex
t
←
L
am
in
atio
n
Step
7
: r
etu
r
n
tex
t
}
Alg
o
r
ith
m
2
.
Pro
p
o
s
ed
B
i
-
L
STM
m
o
d
el
I
n
p
u
t: Pr
ep
r
o
ce
s
s
ed
tex
t
Ou
tp
u
t: C
lass
lab
el
{
Step
1
: Fo
r
to
k
e
n
in
tex
t: //Te
x
t c
o
n
tain
s
a
lis
t o
f
f
ea
tu
r
e
m
a
p
s
Step
2
:
T
ex
tf
←
E
m
b
ed
d
in
g
//
Ap
p
ly
a
B
i
-
L
STM
em
b
ed
d
in
g
lay
er
to
ca
p
tu
r
e
co
n
tex
t
u
al
in
f
o
r
m
atio
n
f
r
o
m
t
h
e
tex
t
Step
3
:
T
ex
tf
←
B
i
-
L
STM
//
P
r
o
ce
s
s
th
e
tex
t
u
s
in
g
a
B
i
-
L
S
T
M
lay
er
to
u
n
d
er
s
tan
d
th
e
s
eq
u
en
ce
an
d
ca
p
tu
r
e
d
ep
en
d
e
n
cies
Step
4
: T
ex
tf
←
Atten
tio
n
//A
p
p
ly
an
atten
tio
n
m
ec
h
an
is
m
t
o
f
o
cu
s
o
n
r
elev
a
n
t p
ar
ts
o
f
th
e
tex
t
Step
5
:
Pl,
Ol
←
Ap
p
ly
s
en
tim
en
t
an
aly
s
is
o
v
er
T
ex
t
f
to
g
et
p
o
lar
ity
an
d
o
b
jectiv
ity
//Us
e
an
ap
p
r
o
p
r
iate
s
en
tim
en
t a
n
aly
s
is
tech
n
iq
u
e
t
o
o
b
tain
t
h
e
s
en
tim
en
t f
ea
tu
r
es
Step
6
:
Pl,
Ol
←
R
an
k
th
e
Pl,
Ol
//R
an
k
th
e
p
o
lar
ity
an
d
o
b
jectiv
ity
s
co
r
es
to
h
ig
h
lig
h
t
s
i
g
n
if
ican
t
s
en
tim
en
t
f
ea
tu
r
es
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ad
v
Ap
p
l Sci
I
SS
N:
2252
-
8
8
1
4
S
en
timen
t a
n
a
lysi
s
o
f v
a
cc
in
e
d
a
ta
u
s
in
g
en
h
a
n
ce
d
d
ee
p
lea
r
n
in
g
a
l
g
o
r
ith
ms
(
Mo
n
ika
V
erma
)
569
Step
7
:
C
t
←
E
x
tr
a
ct
co
n
tex
t
f
r
o
m
tex
t
//Us
e
co
n
tex
t
ex
tr
a
ctio
n
tech
n
i
q
u
es
to
u
n
d
e
r
s
tan
d
th
e
s
ce
n
a
r
io
a
n
d
to
p
ic
o
f
th
e
tex
t
Step
8
:
L
ab
el
←
B
i
-
L
STM
cl
ass
if
ier
m
ak
es
clas
s
if
icatio
n
o
v
er
tex
t
b
y
r
ec
eiv
in
g
Pl,
Ol,
an
d
C
t
as
in
p
u
t
//Us
e
a
B
i
-
L
STM
-
b
ased
class
if
ier
th
at
co
n
s
id
er
s
s
en
tim
en
t a
n
d
c
o
n
tex
t f
ea
tu
r
es f
o
r
tex
t c
lass
if
icatio
n
r
etu
r
n
lab
el
}
Alg
o
r
ith
m
3
.
Pro
p
o
s
ed
f
in
al
B
i
-
L
STM
in
-
d
ep
th
m
o
d
el
Step
1
.
Pre
p
r
o
ce
s
s
in
g
:
I
n
p
u
t: R
aw
tex
t d
ata
ab
o
u
t v
ac
cin
atio
n
.
-
Go
al:
C
lean
an
d
p
r
e
p
ar
e
tex
t f
o
r
an
aly
s
is
.
-
Pro
ce
s
s
:
-
R
em
o
v
e
n
o
is
e
(
e.
g
.
,
UR
L
s
,
s
p
ec
ial
ch
ar
ac
ter
s
)
.
-
No
r
m
alize
tex
t (
e.
g
.
,
lo
wer
ca
s
e,
s
tem
m
in
g
)
.
Step
2
.
E
m
b
e
d
d
in
g
:
I
n
p
u
t: Pr
ep
r
o
ce
s
s
ed
tex
t.
-
Go
al:
C
o
n
v
er
t te
x
t in
to
n
u
m
er
ical
f
o
r
m
at.
-
Pro
ce
s
s
:
-
Use w
o
r
d
em
b
ed
d
in
g
s
lik
e
Gl
o
Ve
to
tr
an
s
f
o
r
m
ea
ch
t
o
k
en
i
n
to
a
d
e
n
s
e
v
ec
to
r
.
-
E
q
u
ati
on:
⃗
⃗
=
(
)
,
ℎ
⃗
⃗
ℎ
.
Step
3
.
B
i
-
L
STM
lay
er
:
I
n
p
u
t: Seq
u
e
n
ce
o
f
wo
r
d
v
ec
to
r
s
.
-
Go
al:
C
ap
tu
r
e
co
n
tex
t a
n
d
d
e
p
en
d
en
cies in
th
e
tex
t.
-
Pro
ce
s
s
:
-
Pas
s
th
e
em
b
ed
d
in
g
s
th
r
o
u
g
h
a
B
i
-
L
STM
lay
er
.
-
E
q
u
atio
n
:
-
Fo
r
war
d
p
ass
:
ℎ
⃗
⃗
⃗
=
(
⃗
⃗
⃗
⃗
,
ℎ
−
1
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
)
-
B
ac
k
war
d
p
ass
:
ℎ
⃖
⃗
⃗
⃗
=
(
⃗
⃗
⃗
⃗
,
ℎ
+
1
⃖
⃗
⃗
⃗
⃗
⃗
⃗
⃗
⃗
)
-
C
o
m
b
in
ed
s
tate:
ℎ
=
[
ℎ
⃗
⃗
⃗
;
ℎ
⃖
⃗
⃗
⃗
]
Step
4
.
Atten
tio
n
m
ec
h
an
is
m
:
I
n
p
u
t: Hid
d
e
n
s
tates f
r
o
m
B
i
-
L
STM
.
-
Go
al:
Hig
h
lig
h
t im
p
o
r
tan
t p
ar
t
s
o
f
th
e
tex
t.
-
Pro
ce
s
s
:
-
Ap
p
ly
an
atten
tio
n
m
ec
h
a
n
is
m
to
th
e
s
eq
u
en
ce
o
f
h
id
d
e
n
s
tates.
-
E
q
u
atio
n
:
∝
=
ex
p
(
(
ℎ
)
)
∑
ex
p
(
(
ℎ
)
)
,
wh
er
e
th
e
s
co
r
e
is
a
f
u
n
ctio
n
m
ea
s
u
r
in
g
th
e
i
m
p
o
r
tan
ce
o
f
ea
ch
h
id
d
en
s
tate.
Step
5
.
Sen
tim
en
t
an
aly
s
is
(
p
o
lar
ity
an
d
o
b
jectiv
ity
)
:
I
n
p
u
t: T
e
x
t f
ea
tu
r
es f
r
o
m
th
e
a
tten
tio
n
m
ec
h
an
is
m
.
-
Go
al:
Dete
r
m
in
e
s
en
tim
en
t p
o
lar
ity
an
d
o
b
jectiv
ity
.
-
Pro
ce
s
s
:
-
Use sen
tim
en
t a
n
aly
s
is
to
o
ls
o
r
m
o
d
els to
p
r
ed
ict
s
en
tim
en
t
s
co
r
es.
-
E
q
u
atio
n
: Pl=Sen
tim
en
tMo
d
el
(
ℎ
)
,
O1
=
O
b
jectiv
itMo
d
el
(
ℎ
)
Step
6
.
C
o
n
te
x
t
ex
tr
ac
tio
n
:
I
n
p
u
t: Pr
ep
r
o
ce
s
s
ed
tex
t.
-
Go
al:
Un
d
er
s
tan
d
th
e
c
o
n
tex
t
an
d
to
p
ic
o
f
th
e
te
x
t.
-
Pro
ce
s
s
:
-
Ap
p
ly
NL
P tec
h
n
iq
u
es to
ex
tr
ac
t th
em
atic
elem
en
ts
.
-
E
q
u
atio
n
:
=Co
n
tex
tEx
tr
ac
to
r
(
t
ex
t
)
Step
7
.
C
lass
if
icatio
n
:
I
n
p
u
t: Sen
tim
en
t sco
r
es (
Pl,
O
l)
an
d
co
n
tex
t f
ea
tu
r
es (
C
t)
.
-
Go
al:
C
las
s
if
y
th
e
s
en
tim
en
t o
f
th
e
tex
t r
e
g
ar
d
in
g
v
ac
cin
atio
n
.
-
Pro
ce
s
s
:
-
I
n
p
u
t th
e
f
ea
tu
r
es in
to
a
B
i
-
L
STM
class
if
ier
.
-
E
q
u
atio
n
: L
a
b
el=
B
i
-
L
STM
C
l
ass
if
ier
(
,
,
)
Step
8
.
Ou
tp
u
t:
Ou
tp
u
t: S
en
tim
en
t la
b
el
(
Po
s
itiv
e,
Neg
ativ
e)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
8
1
4
I
n
t J Ad
v
Ap
p
l Sci
,
Vo
l.
14
,
No
.
2
,
J
u
n
e
2
0
2
5
:
5
6
2
-
579
570
4.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
4
.
1
.
Da
t
a
s
et
T
h
e
d
ataset
h
as
r
ec
en
t
twee
ts
ab
o
u
t
th
e
C
OVI
D
-
1
9
v
ac
cin
es
u
s
ed
in
th
e
en
tire
wo
r
ld
o
n
a
lar
g
e
s
ca
le,
as
f
o
llo
w
s
:
Pfiz
er
/B
io
NT
ec
h
,
Sin
o
p
h
ar
m
,
Sin
o
v
ac
,
Mo
d
er
n
a
,
Ox
f
o
r
d
/As
tr
aZ
en
ec
a
,
Co
v
ax
in
,
an
d
Sp
u
tn
ik
V.
I
n
itial
d
ata
was
m
er
g
ed
f
r
o
m
twee
ts
ab
o
u
t
Pfiz
er
/B
io
NT
ec
h
v
ac
cin
e.
I
ad
d
ed
th
en
twee
ts
f
r
o
m
Sin
o
p
h
ar
m
,
Sin
o
v
ac
(
b
o
th
C
h
in
ese
-
p
r
o
d
u
ce
d
v
ac
cin
es),
Mo
d
er
n
a,
Ox
f
o
r
d
/As
tr
a
-
Z
e
n
ec
a,
C
o
v
ax
in
,
an
d
Sp
u
tn
ik
V
v
ac
cin
es.
T
h
e
co
lle
ctio
n
was
f
ir
s
t
d
ay
s
twice
a
d
a
y
u
n
til
I
i
d
en
tifie
d
a
p
p
r
o
x
im
at
ely
th
e
n
ew
twee
ts
q
u
o
ta
,
a
n
d
th
e
n
co
llectio
n
f
o
r
all
v
ac
cin
es st
ab
ilized
at
o
n
ce
a
d
ay
,
d
u
r
in
g
m
o
r
n
in
g
h
o
u
r
s
(
GM
T
)
.
F
ig
u
r
e
2
d
is
p
lay
s
th
e
d
is
tr
ib
u
t
io
n
o
f
d
i
f
f
er
en
t
s
en
tim
en
t
lab
els
with
in
a
tr
ain
in
g
d
ataset,
s
h
o
wca
s
in
g
th
e
n
u
m
b
er
o
f
s
am
p
les
f
o
r
e
ac
h
s
en
tim
en
t
ca
teg
o
r
y
.
T
h
e
lab
els
in
clu
d
e
"
s
tr
o
n
g
ly
n
e
g
ativ
e,
"
"n
eg
ativ
e,
"
"n
eu
tr
al,
"
"
p
o
s
itiv
e,
"
a
n
d
"str
o
n
g
ly
p
o
s
itiv
e
.
"
T
h
e
"
n
eu
tr
al
"
ca
teg
o
r
y
h
as
th
e
h
ig
h
est
r
e
p
r
esen
tatio
n
,
with
a
s
ig
n
if
ican
t
n
u
m
b
e
r
o
f
s
am
p
le
s
,
s
u
g
g
esti
n
g
a
lar
g
e
p
o
r
tio
n
o
f
th
e
d
ataset
co
n
s
is
ts
o
f
n
eu
tr
al
s
en
tim
en
ts
.
T
h
e
"
p
o
s
itiv
e
"
lab
el
f
o
llo
ws
as
th
e
s
ec
o
n
d
m
o
s
t
f
r
eq
u
e
n
t
ca
teg
o
r
y
,
in
d
icatin
g
a
co
n
s
id
er
a
b
le
n
u
m
b
er
o
f
s
am
p
les
with
p
o
s
itiv
e
s
en
tim
en
t.
T
h
e
"
n
eg
ativ
e
"
lab
el
h
as
a
m
o
d
er
ate
n
u
m
b
e
r
o
f
s
am
p
les,
wh
ile
"
s
tr
o
n
g
ly
p
o
s
itiv
e
"
h
as
f
ewe
r
s
am
p
les,
an
d
"
s
tr
o
n
g
ly
n
eg
ativ
e
"
h
as
th
e
least
n
u
m
b
er
o
f
s
am
p
les,
in
d
icatin
g
a
r
elativ
ely
s
m
all
r
ep
r
esen
tatio
n
o
f
s
tr
o
n
g
n
eg
ativ
e
s
en
tim
en
t
in
th
e
d
atase
t.
T
h
is
d
is
tr
ib
u
tio
n
s
u
g
g
ests
th
at
th
e
d
ataset
is
s
k
ewe
d
to
war
d
s
n
eu
tr
al
an
d
p
o
s
iti
v
e
s
en
tim
en
ts
,
wh
ich
c
o
u
ld
i
n
f
lu
en
ce
th
e
m
o
d
el'
s
p
er
f
o
r
m
a
n
ce
d
u
r
in
g
tr
ain
in
g
,
p
o
ten
tially
f
av
o
r
in
g
t
h
ese
m
o
r
e
f
r
eq
u
e
n
t c
ateg
o
r
ies.
Fig
u
r
e
2
.
Nu
m
b
er
o
f
s
am
p
les i
n
th
e
tr
ain
in
g
s
et
F
ig
u
r
e
3
co
r
r
elatio
n
m
atr
i
x
d
i
s
p
lay
ed
in
th
e
h
ea
tm
a
p
v
is
u
al
izes
th
e
r
elatio
n
s
h
ip
s
b
etwe
en
d
if
f
er
en
t
f
ea
tu
r
es
in
a
d
ataset,
in
clu
d
in
g
id
,
u
s
er
_
f
o
llo
wer
s
,
u
s
er
_
f
r
ie
n
d
s
,
u
s
er
_
f
av
o
u
r
ites
,
u
s
er
_
v
er
if
ied
,
r
etwe
ets,
an
d
f
av
o
r
ites
.
T
h
e
co
lo
r
g
r
ad
ien
t
r
an
g
es
f
r
o
m
d
a
r
k
p
u
r
p
le
in
d
ica
tin
g
a
lo
w
o
r
n
o
co
r
r
elatio
n
,
clo
s
er
to
0
to
b
r
ig
h
t
y
ello
w
in
d
icatin
g
a
h
ig
h
co
r
r
elatio
n
,
clo
s
er
to
1
.
No
tab
l
y
,
th
e
m
atr
ix
s
h
o
ws
a
s
tr
o
n
g
p
o
s
itiv
e
co
r
r
elatio
n
b
etwe
en
r
etwe
ets
an
d
f
a
v
o
r
it
es,
h
ig
h
lig
h
ted
b
y
a
b
r
i
g
h
t
y
ello
w
co
lo
r
,
s
u
g
g
esti
n
g
t
h
at
twee
ts
with
m
o
r
e
r
etwe
ets
ten
d
to
h
av
e
m
o
r
e
f
av
o
r
ites
.
T
h
er
e
is
also
a
m
o
d
er
ate
co
r
r
elatio
n
b
etwe
en
u
s
er
_
f
o
llo
wer
s
an
d
u
s
er
_
f
av
o
u
r
ites
,
as
in
d
icate
d
b
y
a
lig
h
ter
s
h
ad
e
o
f
b
lu
e
,
im
p
l
y
in
g
th
at
u
s
er
s
with
m
o
r
e
f
o
ll
o
wer
s
ar
e
lik
ely
to
h
av
e
m
o
r
e
f
av
o
r
ites
.
Ho
wev
er
,
m
o
s
t
o
t
h
er
c
o
r
r
elatio
n
s
a
p
p
e
ar
wea
k
o
r
n
o
n
-
ex
is
ten
t,
as
s
h
o
wn
b
y
th
e
d
ar
k
er
p
u
r
p
le
s
h
ad
es.
Fo
r
ex
am
p
le,
th
e
id
f
ea
tu
r
e
h
as
alm
o
s
t
n
o
co
r
r
elatio
n
with
an
y
o
th
e
r
v
ar
i
ab
les.
Ad
d
itio
n
ally
,
u
s
er
_
v
er
if
ied
d
o
es
n
o
t
s
h
o
w
a
s
tr
o
n
g
c
o
r
r
elatio
n
with
o
th
e
r
a
ttrib
u
tes,
in
d
icatin
g
t
h
at
v
er
if
icatio
n
s
tatu
s
d
o
es
n
o
t
s
ig
n
if
ican
tly
im
p
ac
t
th
e
n
u
m
b
er
o
f
f
o
llo
wer
s
,
f
r
ien
d
s
,
f
av
o
r
ites
,
o
r
r
etwe
ets.
T
h
is
m
atr
ix
p
r
o
v
id
es
a
co
m
p
r
eh
e
n
s
iv
e
o
v
e
r
v
iew
o
f
h
o
w
th
ese
v
ar
iab
les ar
e
in
ter
r
elate
d
with
in
th
e
d
ataset.
F
ig
u
r
e
4
co
n
f
u
s
io
n
m
atr
ix
il
lu
s
tr
ates
th
e
p
er
f
o
r
m
an
ce
o
f
a
m
u
lti
-
class
cla
s
s
if
icat
io
n
m
o
d
el
b
y
co
m
p
ar
in
g
th
e
tr
u
e
lab
els
a
g
ain
s
t
th
e
p
r
ed
icted
lab
els
ac
r
o
s
s
f
iv
e
s
en
tim
en
t
ca
teg
o
r
ies,
r
e
p
r
esen
ted
b
y
in
d
ices
0
to
4
.
T
h
e
d
iag
o
n
al
elem
en
ts
f
r
o
m
to
p
lef
t to
b
o
tto
m
r
ig
h
t in
d
i
ca
te
th
e
n
u
m
b
er
o
f
co
r
r
ec
t
p
r
ed
ictio
n
s
f
o
r
ea
c
h
ca
teg
o
r
y
:
1
1
,
4
5
4
f
o
r
class
0
,
3
9
,
9
2
9
f
o
r
class
1
,
2
2
,
9
2
4
f
o
r
class
2
,
1
,
0
7
5
f
o
r
class
3
,
a
n
d
3
,
0
4
6
f
o
r
class
4
,
r
ef
lectin
g
s
tr
o
n
g
m
o
d
el
ac
c
u
r
a
cy
in
t
h
ese
ca
teg
o
r
ies.
Of
f
-
d
ia
g
o
n
al
elem
e
n
ts
r
ep
r
esen
t
m
is
class
if
icatio
n
s
,
s
u
ch
as
4
5
0
s
am
p
les
o
f
class
0
in
co
r
r
ec
tly
p
r
ed
icted
as
class
2
,
an
d
4
6
2
s
am
p
les
o
f
class
4
m
is
cl
ass
if
ied
as
clas
s
2
.
T
h
e
m
atr
ix
s
h
o
ws
r
elativ
el
y
h
ig
h
ac
cu
r
ac
y
f
o
r
class
es
1
an
d
2
,
with
m
in
im
al
m
is
class
if
icatio
n
s
ac
r
o
s
s
o
th
e
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ad
v
Ap
p
l Sci
I
SS
N:
2252
-
8
8
1
4
S
en
timen
t a
n
a
lysi
s
o
f v
a
cc
in
e
d
a
ta
u
s
in
g
en
h
a
n
ce
d
d
ee
p
lea
r
n
in
g
a
l
g
o
r
ith
ms
(
Mo
n
ika
V
erma
)
571
class
es.
T
h
e
m
o
s
t
f
r
e
q
u
en
t
m
i
s
class
if
icatio
n
s
o
cc
u
r
in
a
d
jac
en
t
class
es,
in
d
icatin
g
t
h
at
th
e
m
o
d
el
s
o
m
etim
es
co
n
f
u
s
es
s
im
ilar
s
en
tim
en
t
ca
teg
o
r
ies,
s
u
ch
as
m
is
tak
in
g
"
n
eu
tr
al
"
(
class
2
)
f
o
r
"
p
o
s
itiv
e
"
(
class
1
)
o
r
v
ice
v
er
s
a.
Ov
er
all,
th
e
co
n
f
u
s
io
n
m
atr
ix
r
ev
ea
ls
th
at
th
e
m
o
d
el
p
er
f
o
r
m
s
well,
with
th
e
m
ajo
r
ity
o
f
p
r
e
d
ictio
n
s
alig
n
in
g
with
th
e
tr
u
e
la
b
els,
alth
o
u
g
h
th
er
e
a
r
e
ar
ea
s
wh
er
e
f
u
r
th
er
r
ef
in
em
e
n
t
co
u
ld
r
e
d
u
c
e
m
is
class
if
icatio
n
r
ates.
Fig
u
r
e
3
.
C
o
r
r
elatio
n
m
atr
ix
Fig
u
r
e
4
.
L
STM
co
n
f
u
s
io
n
m
a
tr
ix
F
ig
u
r
e
5
co
n
f
u
s
io
n
m
atr
ix
p
r
o
v
id
es
a
d
etailed
e
v
alu
atio
n
o
f
th
e
p
er
f
o
r
m
an
ce
o
f
a
m
u
lti
-
class
class
if
icatio
n
m
o
d
el
b
y
co
m
p
a
r
in
g
ac
tu
al
v
er
s
u
s
p
r
ed
icted
lab
els
ac
r
o
s
s
f
iv
e
s
en
tim
en
t
ca
te
g
o
r
ies,
d
en
o
ted
b
y
in
d
ices
0
to
4
.
T
h
e
d
iag
o
n
al
en
tr
ies
r
ep
r
esen
t
th
e
co
r
r
ec
tly
c
lass
if
ied
in
s
tan
ce
s
f
o
r
ea
ch
ca
teg
o
r
y
:
8
,
6
8
5
f
o
r
class
0
,
2
9
,
9
3
6
f
o
r
class
1
,
1
7
,
2
2
7
f
o
r
class
2
,
7
7
6
f
o
r
class
3
,
an
d
2
,
3
6
3
f
o
r
class
4
,
in
d
ic
atin
g
th
at
th
e
m
o
d
el
p
er
f
o
r
m
s
well
o
v
er
all
with
a
h
ig
h
n
u
m
b
er
o
f
co
r
r
ec
t
p
r
e
d
ictio
n
s
in
ea
ch
ca
teg
o
r
y
.
O
f
f
-
d
iag
o
n
al
en
tr
ies
h
ig
h
lig
h
t
th
e
m
is
class
if
icatio
n
s
,
s
u
ch
as
2
3
3
s
am
p
les
o
f
class
0
in
co
r
r
ec
tly
p
r
ed
icted
as
class
2
an
d
2
7
8
s
am
p
les
o
f
class
4
m
is
cla
s
s
if
i
ed
as
class
2
.
T
h
e
m
o
s
t
f
r
eq
u
en
t
m
is
class
if
icatio
n
s
o
cc
u
r
b
etwe
en
s
im
ilar
o
r
ad
jace
n
t
s
en
tim
en
t
class
es,
lik
e
"
n
eu
t
r
al
"
(
class
2
)
b
ein
g
c
o
n
f
u
s
ed
with
"
p
o
s
itiv
e
"
(
class
1
)
a
n
d
v
ice
v
er
s
a.
T
h
e
m
atr
ix
in
d
icate
s
th
at
wh
ile
th
e
m
o
d
el
g
e
n
er
ally
ac
h
iev
e
s
g
o
o
d
class
if
icatio
n
p
er
f
o
r
m
a
n
ce
,
p
ar
ticu
lar
ly
f
o
r
th
e
m
o
r
e
p
o
p
u
lated
class
es,
th
er
e
ar
e
ce
r
tain
ar
ea
s
wh
er
e
co
n
f
u
s
io
n
b
etwe
en
ad
jace
n
t
s
en
t
im
en
t
class
es
co
u
ld
b
e
r
ed
u
ce
d
to
f
u
r
th
e
r
im
p
r
o
v
e
th
e
m
o
d
el’
s
ac
cu
r
ac
y
an
d
r
elia
b
ilit
y
.
Evaluation Warning : The document was created with Spire.PDF for Python.