Database Reference
In-Depth Information
A
on the topic of document
d
, we propose to compare the quantity of information
he has imported via his other publications. From the information producer's
point of view, this can be measured by the information entropy
H
d
ð
In order to estimate the knowledge and the experience of a coauthor
a
k
2
t
i
Þ
for tags
assigned to the subset of the coauthor publications noted
A
k
. We consider as a
T
d
assigned to document
d
where it exists an edge
random variable each tag
t
i
2
D
) and we calculate its probability distribution Pr
k
(
t
i
) among the
subcollection of documents
A
e
(
t
i
,
d
j
)
2
(
T
¼
S
k¼
1
A
k
published by the
m
coauthors of the
document
d
. The final
Authorship
weight
w
(
a
k
,
d
) is computed as follows:
1
T
d
1
A
kk
y;
H
d
ð
w
ð
a
k
;
d
Þ¼
1
t
i
Þ
(6.3)
kk
where
X
H
d
ð
Pr
k
log Pr
k
t
i
Þ¼
ð
t
i
Þ
ð
t
i
Þ
;
(6.4)
t
i
2
T
d
and
tf
ð
t
i
; A
k
Þ
Pr
k
ð
t
i
Þ¼
0
5
t
i
; A
Þ
þ
0
5
(6.5)
:
:
:
tf
ð
ð
t
i
; A
k
Þ
is the frequency of tag
t
i
in the subset of author
a
k
documents
With tf
ð
A
k
Þ
represents the tag frequency in the subcollection of the
coauthors documents
ð
A
Þ
. In order to get ascendant values of entropy, we
scale tag probability into the interval [0.5, 1]. We note that 1
and tf
ð
t
i
; A
Þ
is the default
weight value attributed to authors having a single document in the collection.
y
Some social network analysis algorithms do not support multiple edges between
two nodes with similar directions. Thus, we propose to combine the coauthorship
and citation weights as follows:
1
4
w
ð
i
j
Þ¼
ð
1
þ
Co
ð
i
j
Þ
Þ
ð
1
þ
Ci
ð
i
j
Þ
Þ
:
(6.6)
;
;
;
6.5.2 Computing Social Relevance
The objective of document relevance estimation within the social network is to
derive a more accurate response for the user by combining the topical relevance of
document
d
and the importance of associated authors in the social network.
Accordingly, we aim to select the social importance measures that identify central
Search WWH ::
Custom Search