Research Challenges of
             Dynamic Socio-Semantic Networks

                                  Rostislav Yavorskiy

                          Witology, Models and Algorithms
                         Kapranova str. 3, Moscow, RUSSIA
                        Rostislav.Yavorskiy@witology.com
                             http://www.witology.com


        Abstract. A general model of a socio-semantic network is presented in
        terms of state-transition systems. We provide some examples and indi-
        cate research directions, which seem to us the most important from the
        application point of view.

        Keywords: social network, semantic network, socio-semantic network


1     Model of a socio-semantic network
1.1     Social network
A social network is usually modeled as a weighted multi-graph
                         G = {V, E1 , . . . , Ek ; π, δ1 , . . . , δk },
where
 – V represents members of the network,
 – E1 , . . . , Ek ⊂ V × V denote different relations between the members, e.g.
   being a friend, follower, relative, co-worker etc.
 – π : V → Π is a user profile function, which stores personal information
   about the network members.
 – δi : Ei → ∆i (i ∈ {1, . . . , k}) keeps parameters and details of the corre-
   sponding relation.

1.2     Content
The model of the content has a very similar definition. It is a multi-graph
                        C = {T, R1 , . . . , Rm ; θ, γ1 , . . . , γm },
where
 – T stands for the set of all elements of the generated content, e.g. posts,
   comments, evaluations, tags etc.
 – R1 , . . . , Rm ⊂ T × T denote different relations on the content, e.g. being a
   reply on, have the same subject, etc.
 – θ : T → Θ stores parameters of the content;
 – γi : Ri → Γi (i ∈ {1, . . . , k}), similarly, keeps parameters and details of the
   corresponding relation.
1.3   Authorship and other relations between the users and the
      content
The basic connections between the social graph and the content are defined by
the authorship relation A,
                                A ⊂ V × T.
One can also consider other kinds of connections of this kind, but usually all of
them could be modeled via introducing a new type of content. For example, the
relation John is interested in post “Announcement” could be modeled by intro-
ducing a new content node interest evidence, which points to “Announcement”
(use the corresponding relation Ri here) and is authored by John.

1.4   The context
Before we turn to description of the socio-semantic network dynamics there is
one more important parameter not to be missed. It is external context, Ω, which
may include different parameters like project or campaign phase, flag for a bank
holiday, or a maintenance status of the network.

2     The socio-semantic network dynamics
Now, when all the components of the network have been defined, the list of pos-
sible system updates, which determine the network evolution, is rather evident:
 – addition of new members to V ;
 – changes in user profiles π;
 – updates of social relations E1 , . . . , Ek and their parameters δ1 , . . . , δk ;
 – creation and update of content nodes in T , (also affects the authorship rela-
   tion A);
 – changes in properties of and relations between the content nodes θ, γ1 , . . . , γm ;
 – changes in context Ω.

3     Examples
3.1   School or a training center
A training center usually has a standardized set of reading materials, textbooks,
tasks, assignments and exam tests. At the same time, the students’ network
evolves permanently. In terms of the definition above one can say that the content
part of this socio-semantic network is rather stable while the social network is
very dynamic.

3.2   Research or analytic team
This example resides on the opposite side of the spectra. The team (the so-
cial network part) is rather stable while the content is actively processed and
generated.
3.3   Fixed term project

A targeted crowdsourcing project or a collective intelligence venture provide an
example of a dynamic socio-semantic network, which is created from the scratch
and is aimed at solving a particular task or a problem. New content is generated
and new members join the network at all stages of its lifecycle.


4     Research challenges
Assume that we have all necessary data about the network dynamics available.
In all the examples mentioned above one can identify two principal tasks for
analysis:

 – Given all the data about the users activities and the content discover the
   right people (knowing, capable, skillfull etc.)
 – Given all the data about the social network dynamics and the content evo-
   lution discover the right texts (interesting, influential, prominent etc.)
    Many promising approaches and useful algorithms have already been devel-
oped during the last decades [1–4], several new ideas are implemented in the
Witology platform [5]. Some have proved to be quite efficient, although most of
them are based on fairly simple mathematical tools. Still, the field is rather in
its rudimentary phase. We believe that the next breakthrough lies in interdis-
ciplinary research covering sociology, psychology, linguistics and other related
fields.


References
1. Sergey Brin, Lawrence Page. The anatomy of a large-scale hypertextual Web search
   engine. Computer Networks and ISDN Systems 30: 107117, (1998).
2. Damon Horowitz, Sepandar D. Kamvar. The Anatomy of a LargeScale Social Search
   Engine. Proceedings of WWW’2010, April 26–30, 2010, Raleigh, North Carolina.
3. Camille Roth, Jean-Philippe Cointet. Social and Semantic Coevolution in Knowl-
   edge Networks, Social Networks, 32(1):16-29 (2009)
4. Jean-Philippe Cointet, Camille Roth. Socio-semantic Dynamics in a Blog Net-
   work, IEEE SocialCom International Conference on Social Computing, Vancouver,
   Canada, August 2009.
5. Witology, “Search of good ideas through search of people, and search of right people
   through search of ideas”, http://www.witology.com