Image Processing Reference
In-Depth Information
social relationships and identify potential communities, while building a social
profile of both individuals and communities. This will create overlay social network
links and relations, based on pure social semantic primitives. This social network
can be further analysed [
11
,
12
,
16
] in order to:
• Detect the social communities with interacting session patterns.
• Build a social profile for the individual user: social connectivity, interaction
topics, number of communities the user belongs to, etc.
• Build a social profile for communities: number of members, group cohesion, etc.
• Profile the communities across a number of attributes (content access, session
primitives, socio-demographics
) based on the interrelation principles that
...
appear in social groups.
By projecting social communities to the network paradigm and narrowing the
social interaction to a more system-scoped level (enabled by technologies like [
17
]
and [
18
]), community formulation could be focusing on users accessing the overlay
through common session topics with the following relation criteria:
• Sharing audio streams
• Sharing 2D or 3D video streams
• Sharing live streams through interaction and collocation in virtual worlds
In order to better explain the issue, let us take as an example the network topology
shown in Fig.
12.2
. Blue nodes are end users and grey nodes are network nodes.
Let us assume that end-user A aims to communicate with end-user D. Assuming
that all network nodes and links are equivalent, from a networking point of view,
the best path is:
A
$
1
$
2
$
6
$
9
$
D
s assume that B and C are very close friends of A and/or D, and based
on the recent history of the social interaction, there is a high probability that B
and/or C will join the session in few minutes. In that case, maybe instead of the
previous path, it is better to select the path:
Now, let
'
A
$
1
$
5
$
7
$
8
$
9
$
D
Though this path is suboptimal, and it is longer by one link as compared to the
previous one, it is at the end a better path, if B and/or C join the session, as content
that already streams via, or is cached at nodes 5 and 7, may be directly multicast or
be forwarded to clients B and/or C via nodes 4 and 10 respectively. However, we
also need to take into account mobility, as if clients B and/or C have a high mobility
profile, they may join from other nodes, reducing the overall benefit of the proposed
path selection.
We argue that the above challenge is facilitated by user community/cluster
detection through social network analysis. Assuming limited nomadic user patterns,
these node clusters can then be mapped to the network topology. The next step is to
