1.3. FAKE NEWS IN SOCIAL MEDIA AGE 7
1.3.5 CHALLENGING PROBLEMS OF FAKE NEWS DETECTION
Organization
Notations
A
w i; j A A
ij
8 1. INTRODUCTION
Notation Meaning
A = {a
1
, ... , a
N
}
the set of news articles
a
i
the embedding vector of news a
i
Y = {y
1
, ... , y
N
}
the set of labels for corresponding news pieces
P = {p
1
, ... , p
K
}
the set of news publishers
S = {s
1
, ... , s
L
}
the set of sentences for a news piece
s
i
= {w
i1
, ... , w
iMi
}
the set of words in a sentence s
i
W
i
the embedding vector for word w
i
I = {i
1
, ... , i
M
}
the set of images for a news piece
E = {e
it
,}
the set of social engagements
{u
i
, c
i
, t} each engagement representing that a user u
i
spreads news article a
using c
i
at time t
U = {u
1
, ... , u
n
}
the set of users engaged in the spreading for a news piece
u
i
the embedding vector of users
C = {c
1
, ... , c
n
}
the set of social media posts in the spreading of a news piece
c
i
the embedding vector of post c
i
c
i
= {w
i1
, ... , w
iQi
}
the set of words in a post c
i
Σ
the word vocabulary of text corpus
X ∈ R
N×d
the news-word matrix with d = |Σ|
X ∈ R
N×d×d
the news-word-word tensor with d = |Σ|
Y ∈ R
N×m×l
the news-user-community tensor
(s, p, o)
the tuple representing a knowledge: a subject s is related to the object
entity by the predicate relation p
λ(t)
the intensity function of temporal point process
E(·)
the empirical expectation function
Beta(α, β) the beta distribution with hyperparameters α and β
A ∈ R
m×m
the user-user adjacency matrix among users engaged in news spreading
U ∈ R
m×k
the user latent representation matrix with latent dimension k
E ∈ R
m×N
the user-news interaction matrix
F ∈ R
N×k
the news latent representation matrix with latent dimension k
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.