THE SNARE LANGUAGE OVERVIEW
Alexandre Barão and Alberto Rodrigues da Silva
IST/INESC-ID, R. Alves Redol, 9, 1000-029 Lisboa, Portugal
Keywords: Social network, Design, Modeling.
Abstract: Social network systems identify existing relations between social entities and provide a set of automatic
inferences on these relations, promoting better interactions and collaborations between these entities.
However, we find that most of existing organizational information systems do not provide, from scratch,
social network features, even though they have to manage somehow social entities. The focus on this paper
starts from this fact, and proposes the SNARE Language as the conceptual framework for SNARE system,
short for “Social Network Analysis and Reengineering Environment”. The SNARE’s purpose is to promote
social network capabilities in information systems not designed originally for the effect. Visual models are
needed to infer and represent new or established patterns of relations. This paper overviews the SNARE
language and shows its applicability through several models regarding the application of the SNARE to the
LinkedIn real scenario.
1 INTRODUCTION
A social network consists of a finite set of actors
and the relations defined among them (Wasserman
and Faust, 1994). Actors are discrete individuals,
corporate or collective social units, and are linked to
one another by social ties (Wasserman and Faust,
1994). A dyad is a linkage or relation between two
actors. Triads are triples of actors and associated
ties. To a large extent, the power of network analysis
lies in the ability to model the relations among
systems of actors. A subgroup of actors is any subset
of actors and all ties among them. A group is the
collection of all actors on which ties are to be
measured. The collection of ties of a specific kind of
members of a group is called a relation (Wasserman
and Faust, 1994). Actors may be referred as social
entities.
An entity is social if involves a network of
relations with other social entities (Masolo et al.,
2004). A social entity play several roles in the same
network. A role is a combination of particular sets of
behavioral, meaningful, and structural attributes
(Welser et al., 2007). The nature of roles and the
way of representing them have been discussed in
different fields, e.g. knowledge representation,
knowledge engineering, object-oriented and
conceptual modeling, multi-agent systems,
linguistics, and cognitive semantics (Masolo et al.,
2004). Four common features about social roles can
be found: (1) roles are properties, e.g. different
entities can play the same role; (2) roles are anti-
rigid and they have dynamic properties, e.g. an
entity can play different roles simultaneously, an
entity can change role, an entity can play the same
role several times, simultaneously, a role can be
played by different entities simultaneously or at
different times, the sequence in which roles may be
acquired and relinquished can be subject to
restrictions; (3) roles have a relational nature, i.e.
roles imply patterns of relations; and (4) roles are
linked to contexts, i.e. a contextual approach refer to
a variety of factors, including relations, events,
organizations and behaviors. The term “context” can
have different interpretations, e.g. metaphysical
context, cognitive context; and linguistic context.
See (Masolo et al., 2004) for a further review.
There are different types of social networks.
One-mode networks involve just a single set of
social entities. Two-mode networks involve two sets
of actors, or one set of actors and one set of events
(Wasserman and Faust, 1994). Events have a time
associated with them and it is possible for relations,
positions and roles to change over time. In spite,
events can occur at different times, the organizers of
events change over time, and a different set of actors
might participate in each event (Licamele et al.,
2005). Dyadic networks and affiliation networks are
344
Barão A. and Rodrigues da Silva A. (2010).
THE SNARE LANGUAGE OVERVIEW.
In Proceedings of the 12th International Conference on Enterprise Information Systems - Information Systems Analysis and Specification, pages
344-349
DOI: 10.5220/0002975203440349
Copyright
c
 SciTePress

particular cases of two-mode networks. Another
kind of network is the ego-centered network where a
focal actor (termed “ego”) has a set of alters who
have ties to ego, and measurements on the ties
among these alters. It is possible to consider three or
more mode networks, but rarely have social network
methods been designed for such data structures
(Wasserman and Faust, 1994).
In Social Network Analysis (SNA) scope,
dynamics of groups are studied to identify relations
and interactions among their members. Starting from
these interactions it is possible to identify social
patterns (Haythornthwaite, 2005) and it is possible
to detect or propose social or organizational changes
that reveal how networks grow or should change.
Also, it is possible to find potential causes and
consequences of a network change, previewing and
controlling networks evolution (Churchill and
Halverson, 2005). These features are dependent of
metrics to allow group properties identification or to
characterize individual influence on a specific group.
Typically scenarios are strategic alliances and
collaborations, flows of information
(communication), affect (friendship), goods and
services (workflow), and influence (advice) (Brass et
al., 2004). Network research represents a different
paradigm of research which requires new concepts
and methods (Borgatti, 2003).
Traditional SNA studies use much information
residing in archives that were not created expressly
for social research. Sometimes, such data provide
measures of social ties and trace relations of social
entities who are reluctant to interviews. Archival
data are often inexpensive, especially when in
electronic form. The validity of archival data rests
on the correspondence between measured
connections and the conceptual ties of research
interest (Carrington et al., 2005). The data
comprising social networks tend to be
heterogeneous, multirelational, and semi-structured.
Link mining is a relevant example showing a
confluence of research in social networks, link
analysis, hypertext and Web mining, graph mining,
relational learning, and inductive logic programming
(Han and Kamber, 2006).
New visual models are needed to infer and
represent patterns of relations, and this paper
proposes the SNARE language as the conceptual
framework for SNARE system. The SNARE system
purpose is to promote social network capabilities in
information systems not designed originally for the
efect.
In Section 1, we introduce social network
concepts. Section 2 overviews social networks
modeling techniques and the motivation for a social
network language. Section 3 purposes the SNARE
language. Finally, Section 4 presents preliminary
conclusions of the investigation.
2 MODELING SOCIAL
NETWORKS
Social network models allow researchers to
conceptualize social structures as patterns of
relations, and understand how an individual is
influenced by a social structural environment
(Wasserman and Faust, 1994). The aim for using
formal methods to show social networks such as
mathematical and graphical techniques is to
represent the descriptions of networks compactly
and systematically. In the analysis of complete
networks, three strategies for modelling social
networks can be found: (1) descriptive methods, also
through graphical representations; (2) mathematical
analysis procedures, often based on a decomposition
of the adjacent matrix; and (3) statistical models
based on probability distributions (Jamali and
Hassan, 2006).
First, graphical representations: graph theory
provides a vocabulary which can be used to label
social structural properties. Also, gives a
representation of a social network as a model
consisting of a set of actors and the ties between
them. When a graph is used as a model of a social
network, points or nodes are used to represent social
entities, and lines, connecting the points, are used to
represent the ties between them. Figure 1 (adapted)
is a graph that describes the structure of relations
between the entities A, B, C, D, E, F and G
(Churchill and Halverson, 2005). In the figure, the
circles are nodes and lines between them are links
(Churchill and Halverson, 2005). The links
corresponds to the sending messages act between
entities. Entity A is linked to two subgroups and also
to an isolated entity G. The arrows shows which are
the directed connections (e.g. A sends mail to E) or
undirected (e.g. A sends mail to F and F sends email
to A) (Churchill and Halverson, 2005). Node A can
be characterized as a boundary entity between two
subgroups, and potentially a point of connection
between them.
The visual representation of data that a graph
offers allows researchers to uncover patterns that
might otherwise go undetected. Graphs have been
widely used in SNA as a mean of formally represent
social relations and quantify social structural
THE SNARE LANGUAGE OVERVIEW
345

Figure 1: “Send Mail To” Graph.
wide review on this approach (Iacobucci, 1994).
Still, inspired by social tagging mechanisms, Peter
Mika has formulated a generic model of semantic-
social networks in the form of a graph of person,
concept and instance associations, extending the
traditional concept of ontologies (concepts and
instances) with the social dimension. His work
showed how community-based semantics emerges
from this model through a process of graph
transformations (Mika, 2005).
Second, mathematical analysis procedures:
matrices are another way to represent networks. A
matrix contains the same information as a graph, but
is more useful for computer analysis, because matrix
operations are widely used for definition and
calculation in SNA (Wasserman and Faust, 1994).
The adjacency matrix is the primary matrix used in
SNA, usually referred as a sociomatrix (Iacobucci,
1994). The entries in the matrix indicate whether
two nodes are adjacent or not. The incidence matrix,
records which lines are incident with which nodes
(Iacobucci, 1994). Figure 2 (adapted) shows a matrix
representing the Figure 1 connections (Churchill and
Halverson, 2005). In the matrix, value 1 indicates
the presence of a connection and value 0 the
absence. The absence of a link between A and D is
represented by the zero value in both cells of the
matrix. Entity A is related to E via a directed link
and E is not directed to A (Churchill and Halverson,
2005).
And third, statistical models: earlier statistical
methods for SNA were introduced by Wasserman
and Faust (Wasserman and Faust, 1994), but in
recent years there has been a growing interest in
exponential random graph models (ERGMs) called
the p* class of models. The exponential random
graph models describe a general probability
distribution of graphs on n nodes. The possible ties
among nodes of a network are regarded as random
variables, and assumptions about dependencies
among these random tie variables determine the
general form of the ERGM for the network. The
Markov random graphs are one particular class of
Figure 2: “Send Mail To” Matrix.
ERGMs (see (Robins et al., 2007) for a summary to
the formulation and application of ERGMs for social
networks). Mathematical and graphical SNA
techniques allow to represent the descriptions of
networks compactly and systematically. For small
populations of actors (e.g. the people in a
neighbourhood, or the business firms in an industry)
it is possible to describe the pattern of social
relations that connect the actors using words.
However, to list all logically possible pairs of actors,
and describe each kind of possible relations, if the
number of actors and number of relation types is
large, formal representations ensure that all the
necessary information is systematically represented,
and provides rules for doing so in ways that are
much more efficient than lists (Hanneman, 2010).
Robert Hanneman considers that social network
analysis is more a branch of "mathematical
sociology” than "statistical or quantitative analysis"
though networkers most certainly practice both
approaches (Hanneman, 2010). He advocates that
the distinction between the two approaches is not
clear. In his words: “Mathematical approaches to
network analysis tend to treat the data as
deterministic. That is, they tend to regard the
measured relationships and relationship strengths as
accurately reflecting the real or final or equilibrium
status of the network. Mathematical types also tend
to assume that the observations are not a sample of
some larger population of possible observations;
rather, the observations are usually regarded as the
population of interest.
Statistical analysts tend to regard the particular
scores on relationship strengths as stochastic or
probabilistic realizations of an underlying true
tendency or probability distribution of relationship
strengths” (Hanneman, 2010). In reviewing the main
results of the analysis and modelling of networks,
Watts describes the main network modelling
approaches, regarding structure, connectivity,
searchability, and degree distributions. He concludes
that the current generation of network-related
research is a rapidly emerging, and a highly
interdisciplinary synthesis occurs, with new
analytical techniques with greater computing power,
and an unprecedented volume of data (Watts, 2004).
ICEIS 2010 - 12th International Conference on Enterprise Information Systems
346

The scope and use of statistical approaches have
been extended in recent years, through methods for
SNA focusing on longitudinal network data, which
is understood as two or more repeated observations
of a graph on a given node set. Longitudinal network
data is the most frequently network data format in
social sciences. Several lectures discuss models
designed to analyze such data, as proposed in
(Snijders, 2001) and (Snijders et al., 2007). There is
little difference between conventional statistical
approaches and SNA approaches. Regarding to
(Hanneman, 2010), univariate, bi-variate, and even
many multivariate descriptive statistical tools are
commonly used to describe, explore, and model
social network data. Social network data is easily
represented as arrays of numbers. For Hanneman,
algorithms from statistics are commonly used to
describe characteristics of individual observations
and the network as a whole, and concludes that
statistical algorithms are very heavy used in
assessing the degree of similarity among social
entities, and if finding patterns in network data (e.g.
factor analysis, cluster analysis, multi-dimensional
scaling). Even the tools of predictive modeling are
commonly applied to network data (e.g. correlation
and regression).The most common emphasis in the
application of inferential statistics to social science
data is to answer questions about the stability,
reproducibility, or generalizability of results
observed in a single sample (Hanneman, 2010).
Using mathematical and graphical SNA
techniques to describe a network, consider a
scenario with multiple relations, i.e. where there
may be more than one relation in a social network
data set (e.g. more than one relation defined on pairs
of actors from group ),  represents the number of
relations. Each relation can be represented as a graph
and has a set of arcs 

containing 

ordered pairs
of actors with 1. Each set  defines a
directed graph with nodes in . These graphs can
be seen in one or more pictures. Each relation is
defined on the same set of nodes, and each has a
different set of arcs. A relation  is given by
 , 

 with 1,2,…,. Consider Figure 3
which presents a scenario for three possible
relations.
We conclude that the methods outlined above are
essential to analyze existing social networks but we
believe that new visual models are needed to infer
and represent new or established patterns of
relations.
Figure 3: Multiple Relations Scenario.
3 SNARE LANGUAGE
The SNARE language proposed in this paper
provides a representation of an abstract social
network structure using UML (www.uml.org) as a
formal descriptive method. SNARE is acronym for
“Social Network Analysis and Reengineering
Environment”. It is an engineering artifact to
represent social networks and allow researchers to
design and build real scenarios for social networks
extraction and relational knowledge discovery. As
mentioned before, this language is the conceptual
root for SNARE system, which has the main purpose
to support social network analysis in information
systems not designed originally for the effect.
Through the instantiation of SNARE language, it is
possible to analyze social entities and multiple
relations among them. Based on dynamic and
multiple aggregations, this language supports N-
mode networks. SNARE language main concepts
are: Social Entity, Relation, Role, Action and Event
as depicted in Figure 4. Three additional concepts to
support multiplicity and give flexibility were
engineered: RelationExtreme, ActionExtreme and
EventExtreme.
Figure 4: SNARE Language Key Concepts.
The SocialEntity represents an entity, typically a
person, an organization, a department or a group in
general terms.
The Relation represents a kind of connection
between two or more social entities and can be
THE SNARE LANGUAGE OVERVIEW
347

expressed in different ways depending on the scope.
In a family context we consider relations such as
IsFatherOf, IsSonOf, IsBrotherOf, etc. In an
enterprise context other relations emerge, for
example, IsColleagueOf, IsBossOf, IsCustomerOf or
IsSupplierOf. In an academic context, we find
relations such as IsTeacherOf, IsMentorOf,
IsStudentOf or IsResearcherWith. The Relation
encapsulates much of the semantic that characterizes
the connection between social entities. However, is
not always a simple task to find the correct semantic
to describe a relation. If we consider the relation
IsColleagueOf, the relation’s semantic definition is
facilitated since the relation is bidirectional. That is,
if the person A is a colleague of the person B, then B
is also colleague of A. On the other hand,
considering the family context again, the father-son
relation, we can not concentrate all the semantics
that characterizes the relation in a single class
Relation. If A is the father of B, we can not say that
B is the father of A. I.e. the relation is not
bidirectional. However, this language supports the
representation of any real-world connection due to
the RelationExtreme concept, that gives flexibility
and characterizes any type of relation. Each entity
has a role. So, the RelationExtreme maintains the
consistency of the connection as it allows
differentiate roles in the same relation.
The Role support semantic roles in a given
context, such as teacher, student, father, child,
administrator or executive director. Returning to the
family context model, the case father-son, probably
the best solution is to define a relation
IsMemberOfFamily and use the roles of each social
entity to differentiate the semantic aspects of father
and son. The SNARE language supports also this
flexibility.
In social relations, it occurs sometimes different
types of actions. The Action concept captures these
flows between entities. The SNARE language makes
it possible to keep track actions performed by social
entities. For example, in an academic context,
considering ResearchesWith relation, WriteAnArticle
can be defined as an Action. Thus, the SNARE
language allows keeping track of all articles written
by participants in the relation: ResearchesWith.
Finally, events are part of people's lives.
Participants of events do not always have met
before. However, when participating in an event,
probably a social entity will get involved in new
relations. In order to accomplish this fact, we decide
to include in our language the concept Event.
The SNARE language ensures that relations,
actions and events can have multiple extreme
instances, this is a flexibility requirement.
4 DISCUSSION
This paper introduces the problems and motivation
behind our research work and overviews the
proposed SNARE language.
Social network analysis is an emergent technique
to identify and understand relations and interactions
among social entities, related patterns and meanings,
to support social or organizational changes that
reveal how networks grow, and find potential causes
and consequences of a network change, previewing
and controlling networks evolution.
In the analysis of complete networks, we found
three strategies for modelling social networks: (1)
descriptive methods, also through graphical
representations; (2) mathematical analysis
procedures, often based on a decomposition of the
adjacent matrix; and (3) statistical models based on
probability distributions. Graphical representations
such as graphs tend to show connections between
social entities ignoring valuable semantic aspects of
the relations. Mathematical and statistical methods
are focused on achieving results which are translated
into algebric expressions, numerical matrices or
coefficients to analyze. We consider that
mathematical analysis procedures and statistical
models complement our work, but our approach is a
different way for graphic representation of social
networks and semantic descriptions. Common
techniques shows social networks as maps of entities
and connections between them. These concepts are
often displayed in a diagram, where nodes are the
points and ties are the lines. There can be many
kinds of ties between the nodes. I.e. there can be
many relation types between social entities, and the
resulting diagrams are often very complex to
uncover related semantic concepts. In order to
understand how an individual is influenced by a
social structural environment, it is also necessary to
identify the semantic of relations in a given social
network. This process helps researchers to
conceptualize and identify social structures as
patterns of relations.
To conceptualize social structures in a network
using our language to model social networks, the
process requires the instantiation of social entities,
roles, relations, actions and events. To do this
instantiation, a set of stereotypes can be used. The
richness of this language to model social networks
ICEIS 2010 - 12th International Conference on Enterprise Information Systems
348

comes from the flexibility to combine these
stereotypes. The flexibility is expressed by all
possible links that may exist on a network without
adding redundant instantiations. E.g. a social entity
can play several roles in the same relation, and this
concept should be achieved through the instantiation
of a factorized Role stereotype. Regarding to the
connection patterns we studied, SNARE language
captures all the possible social network relations.
Also, it is possible to introduce new stereotypes or
adapt existing ones. SNARE language ensures that
relations, actions and events can have multiple
extreme instances and the social network system
keeps references to all previous concepts. After
applying SNARE language to several scenarios, we
conclude that it is flexible to fit the needs of
modeling social networks. Considering
organizational consulting processes, instead of
statistical or mathematical representations, the
notation we use leads to a significant easing of
communication, visualization and discussion. When
comparing with other presented social networks
representation techniques (Figures 1, 2 and 3),
SNARE language includes a new collection of
diagrammatic model elements. These elements are
more expressive to capture social network semantic
concepts. Also, they are unambiguous and supported
by UML tools. The SNARE language notation is a
well-known standard derived, which can grows as
the requirements for modeling grow. If the basic
functionality of SNARE language is not sufficient, it
is possible to extend it through the use of
stereotypes.
From the research discussed in this paper, we
conclude that much work on the area of social
network analysis is still open, and that this area has a
growing potential that should be explored. As a
consequence of this project, we hope to provide new
approaches and technologies to improve social
network analysis for organizational environments. In
the future, our goal is to provide a tool for social
networks patterns design and analysis.
REFERENCES
Borgatti, S. P., The State of Organizational Social
Network Research Today, Boston University, 2003.
Brass, D. J. G., Joseph; Greve, Henrich R.; Tsai, Wenpin,
Taking Stock of Networks and organizations: a
multilevel perspective, in Academy of Management
Journal, 47(6), 2004.
Carrington, P. J., J. Scott, and S. Wasserman, Models and
methods in social network analysis. In Structural
analysis in the social sciences, 27, Cambridge
University Press, 2005.
Churchill, E. and C. Halverson, Social Networks and
Social Networking. IEEE Internet Computing, 9 (5),
2005.
Han, J. and M. Kamber, Data mining : concepts and
techniques. Morgan Kaufmann, 2006.
Haythornthwaite, C., Social Network Methods and
Measures for Examining E-Learning. Graduate
School of Library and Information Science, University
of Illinois at Urbana-Champaign. 2005. Available at:
http://classweb.lis.uiuc.edu/~haythorn/
Iacobucci, D., Graphs and Matrices, in Social Network
Analysis: methods and applications. Structural
analysis in social the social sciences series, C.U.
Press, 1994.
Jamali, M. A. and A., Hassan, Different Aspects of Social
Network Analysis, in Proceedings of the 2006
IEEE/WIC/ACM, 2006.
Licamele, L. B., Mustafa; Getoor, Lise; Roussopoulos,
Nick, Capital and Benefit in Social Networks, in
Proceedings of the LinkKDD, 2005.
Masolo, C. V., Laure; Bottazzi, Emanuele; Catenacci,
Carola; Ferrario, Roberta; Gangemi, Aldo; Guarino,
Nicola. Social Roles and their Descriptions. in
Proceedings of the 9th ICPKRR, AAAI Press. 2004.
Mika, P., Ontologies are us: A unified model of social
networks and semantics, in Proceedings of the 4th
ISWC, LNCS 3729, Springer-Verlag. 2005.
Robins, G. P., Pip; Kalish, Yuval; Lusher, Dean, An
Introduction to exponential random graph (p*) models
for social networks. In Social Networks, 29, Elsevier,
2007.
Wasserman, S. and K. Faust, Social network analysis:
methods and applications. In Structural analysis in the
social sciences, 8, Cambridge University Press. 1994.
Watts, D., The “New Science of Networks”, in Annual
Rev. Sociol. 30, 2004.
Welser, H. T. G., Eric; Fisher, Danyel; Smith, Marc,
Visualizing the Signatures of Social Roles in Online
Discussion Groups, in The Journal of Social Structure.
8 (2), 2007.
Snijders, T. A. B.,
The statistical evaluation of social
network dynamics. In M. E. Sobel and M.P. Becker
(Eds.), Sociological methodology. Basil Blackwell,
2001.
Snijders, T. A. B. et al., Modeling the co-evolution of
networks and behavior. In K. van Montfort, H. Oud, &
A. Satorra (Eds.), Longitudinal models in the
behavioral and related sciences, Lawrence Erlbaum,
2007.
Hanneman, Robert A., Introduction to Social Network
Methods, Riverside, 2010. Available at:
www.analytictech.com/networks.pdf
Object Management Group, UML Website (accessed in
May 2010): www.uml.org.
THE SNARE LANGUAGE OVERVIEW
349