Measuring the dynamic bi-directional influence
between content and social networks
Shenghui Wang and Paul Groth
{swang, pgroth}@few.vu.nl
VU University Amsterdam
De Boelelaan 1081a, 1081 HV, Amsterdam, The Netherlands
Abstract. The Social Semantic Web has begun to provide connections
between users within social networks and the content they produce across
the whole of the Social Web. Thus, the Social Semantic Web provides a
basis to analyze both the communication behavior of users together with
the content of their communication. However, there is little research com-
bining the tools to study communication behaviour and communication
content, namely, social network analysis and content analysis. Further-
more, there is even less work addressing the longitudinal characteris-
tics of such a combination. This paper presents a general framework for
measuring the dynamic bi-directional influence between communication
content and social networks. We apply this framework in two use-cases:
online forum discussions and conference publications. The results pro-
vide a new perspective over the dynamics involving both social networks
and communication content.
1 Introduction
Does an informative post on a microblogging service lead to a user gaining
followers? If a user is popular in a social network, will their new status updates
be widely quoted? If a researcher identifies a new topic one year, does that
result in the research having more coauthors the next? As an increasing amount
of content is mediated through social networks, these types of questions are of
great interest, in particular, to developers, social scientists, and business that
aim to understand the link between content generation and social connection. A
key aspect to answering these questions is to understand how the relationships
between users influence the content of their communication and vice versa.
In this paper, we extend our work in [26] by proposing a general framework
for measuring such influence over time. In our approach, we translate both user
relationships and content into two corresponding networks: a social network and
a content networks. The networks are then characterized using common network
properties such as (in-/out-)degree and betweenness centrality. The influence is
then measured using a set of multilevel time-series regression models producing
what we term an influence network showing how these variables impact each
other in time. Additionally, our Influence Framework can integrate other network
properties tailored to a given problem domain.

The use of the Influence Framework is facilitated by the emergence of Se-
mantic Web technologies not only to represent relationships between users on
the Social Web but also to link to the content those users exchange. For ex-
ample, the Semantically Interlinked Online Communities (SIOC) ontology is for
the representation of the content of discussions but is explicitly intertwined with
the Friend of a Friend (FOAF) ontology that is used to represent personal rela-
tionships. Because the Semantic Web provides these explicit links, it is easier to
obtain the input data sets required by our Influence Framework. Thus, as more
Social Web content is made available using Semantic Web standards, the Frame-
work can be used to investigate a wider variety of content and social networks.
Later, we show how the Influence Framework can be applied to networks ob-
tained by querying the Semantic Web Dog Food dataset [24] as well as networks
extracted from a Dutch political forum. The ability to study the connection be-
tween people through their objects was posited as a key benefit to the Social
Semantic Web [5]. This work is an example of where these benefits are coming
to fruition.
In summary, the contributions of this paper are as follows:
– A general framework for measuring the bi-directional influence between net-
works of people and the content associated with those people.
– A multilevel time-series regression model for measuring the longitudinal in-
fluences between the network properties of content and social networks.
– The generation of influence networks for both Dutch political forums and
the World Wide Web conference series, which provide new material for social
scientists to investigate these domains.
The rest of this paper is organized as follows. We begin by presenting the
Influence Framework and its constituent parts. This is followed by a discussion
of the application of the Framework to two use cases: one studying a conference
series and the other studying data from a Dutch political forum. Related work
is then discussed followed by a conclusion.
2 Influence Framework
The Influence Framework is a three stage framework for measuring the influ-
ence between (and within) user relationships and the content they communicate.
While such measures of influence are clearly possible to perform on a case-by-
case basis, a key realization in this work is that by representing content and
user relationships as networks, standard network properties can provide a good
initial insight into influence in different domains. We note that influence is a
time-dependent notion and thus our framework requires time series data.
The three stages of the framework are:
1. Network Generation
2. Measuring Network Properties
3. Time Series Analysis
We now discuss each of these stages.