Information Technology Reference
In-Depth Information
Trust metrics in a web of trust
For trust metrics in a web of trust studies on trust as a distance which is
calculated from the transitive relationship through the mutually related informa-
tion over the web, Golbeck and Hendler proposed the TrustMail (Golbeck and
Hendler
2004a
,
b
) system.
In their system, each entity fixes its reputation based on the information of other
entities. They call it a 'web of trust.' They are using an ontology to represent
trust and reputation. This enables calculation of the degree of trust between any
two entities. They call it a “trust metric.”
5.4.2 Trust Based on the Reputation Over the Social Web
5.4.2.1 Trust Relationship Based on Follow-Up Relationship of Twitter
Google's PageRank (Brin and Page
1998
) is the most famous method to evaluate
web pages. It is similar to the evaluation method for literature which uses the
number of citations. PageRank evaluates the value of a web page from the number
of hyper links. A lot of methods similar to PageRank have been proposed for the
evaluation of trust of persons in a social web like EigenTrust (Kamvar et al.
2003
).
We have tried to evaluate the trust of a person from the number of followers on
Twitter. We regard the following relationship as a recommendation of a person. We
call this method “PersonaRank” (Yamasaki
2010
).
The features of PersonaRank are: a person that is followed by many people will
be valuable persons, and a person who is followed by a valuable person will also be
a valuable person. The calculation model of PersonaRank is almost the same as
Google PageRank. The outline of the calculation is as follows (Fig.
5.2
).
P
i
:
Persona i
r
ð
P
i
Þ :
The PersonaRank of
P
i
F
P
i
:
The set of followers of
P
i
j
P
i
j :
The number of people following
P
i
H
:
Sub
stochastic following relations matrix
G
:
Google matrix
a
:
lonely node vector
a :
Scaling parameter
;
a scalar between 0 and 1
e
:
Row vector
ð
every element is 1
Þ
T
p
:
PersonaRank vector
p
ðkþ
1
ÞT
¼ p
ðkÞT
G
(5.1)
