4.2. WEAKLY SUPERVISED FAKE NEWS DETECTION 65
F
0
l
1
F
0
min
F
0
RQ
T
2
F
C .kRk
1
C kQk
1
/;
R 2 R
N k
Q 2 R
M k
.kRk
1
C kQk
1
/
4.2.3 A PROBABILISTIC GENERATIVE UNSUPERVISED APPROACH
66 4. CHALLENGING PROBLEMS OF FAKE NEWS DETECTION
(b) Agreeing to the authenticity of the news
(a) Doubting the authenticity of the news
Hierarchical User Engagements
A D fa
1
; ; a
N
g
U
M
U
K
a
i
2 A U
M
i
2 U
M
a
i
u
j
2 U
M i
U
K
ij
2 U
K
News
Verified User
Unverified User
y
y
i
y
N
z
i,
z
i,j
z
i,M
r
i,j,
r
i,j,k
r
i,j,K
… …
… …
4.2. WEAKLY SUPERVISED FAKE NEWS DETECTION 67
u
j
2 U
M
i
z
i;j
2 f0; 1g
z
i;j
z
i;j
a
i
c
i;j
u
j
z
i;j
c
i;j
u
j
a
i
r
i;j;k
2 f0; 1g u
k
2 U
K
ij
u
k
u
k
u
k
r
i;j;k
Probabilistic Graphic Model for Inferring Fake News a
i
z
i;j
r
i;j;k
y
α
0
α
φ
0
ψ
0,0
z
i,j
r
i,j,k
θ
i
Υ
j
k
ψ
0,
k
ψ
,0
β
0,0
β
0,
β
,0
k
ψ
,
β
,
k
φ
j
N
M
K
K
m
news a
i
y
i
i
y
i
Bernoulli.
i
/:
i
D
.
1
;
0
/
i
.
1
;
0
/
1
0
68 4. CHALLENGING PROBLEMS OF FAKE NEWS DETECTION
verified user u
j
1
j
0
j
u
j
1
j
WD p.z
i;j
D 1jy
i
D 1/
0
j
WD p.z
i;j
D 1jy
i
D 0/:
1
j
˛
1
D .˛
1
1
; ˛
1
0
/
˛
1
1
˛
1
0
0
j
˛
0
D .˛
0
1
; ˛
0
0
/
1
j
˛
1
1
; ˛
1
0
0
j
˛
0
1
; ˛
0
0
:
1
j
0
j
u
j
i
x
i
j
y
i;j
Bernoulli.
x
i
j
/
unverified user
u
k
2 U
K
0;0
k
WD p.r
i;j;k
D 1jx
i
D 0; z
i;j
D 0/
0;1
k
WD p.r
i;j;k
D 1jx
i
D 0; z
i;j
D 1/
1;0
k
WD p.r
i;j;k
D 1jx
i
D 1; z
i;j
D 0/
1;1
k
WD p.r
i;j;k
D 1jx
i
D 1; z
i;j
D 1/
.u; v/ 2 f0; 1g
2
u;v
k
u
k
u
v
u;v
k
u;v
k
ˇ
u;v
1
; ˇ
u;v
0
:
y
i
z
i;j
y
i
;z
i ;j
k
r
i;j;k
D
y
i
;z
i;j
k
:
y
Oy D
y
•
p.y; z; r; ; ; / ;
f
0
;
1
g f
0;0
;
0;1
;
1;0
;
1;1
g
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.