Community Detection in Collaborative Tagging Systems - Community-Built Databases: Research and Development - page 108

5.1 Introduction

Collaborative data is nowadays prevalent across the Web. The wide adoption of
technologies that enable users to connect to each other and to contribute to the online
community has revolutionized the way that content is organized and shared on the
Web. In many online applications, it is possible for users to upload their own content
or links to existing content (bookmarking) and to organize it by use of tags, i.e., free-
form keywords. Such applications, examples of which are delicious, 1 flickr, 2 and
BibSonomy, 3 are commonly referred to as Collaborative Tagging Systems.
An established means of modeling the structure of Collaborative Tagging Sys-
tems is the folksonomy model [ 1 , 2 ]. A folksonomy comprises three types of entities
that are of interest in a Collaborative Tagging System: Users ( U ), Resources ( R ), and
Tags ( T ). In addition, it encodes the associations among these entities in the form of a
network, i.e., a tag assignment by a user to a resource is modeled as a set of edges
between the respective nodes (tags, resource, user) on the folksonomy network
under study.
A characteristic property of folksonomy networks, similar to other complex
networks, is the existence of community structure [ 3 ]. The entities of a folksonomy
tend to form groups that are more closely related to each other than to the rest of
the folksonomy entities. For instance, one can identify sets of resources within a
Collaborative Tagging System that are focused on a specific topic. Such community
structure can be explicitly declared when appropriate mechanisms, such as the
Flickr Groups , are provided by the tagging system. Even more interesting is the
implicit community structure that can be discovered by analyzing the network
structure arising from the tagging activities of users.
This chapter presents an overview of research pertaining to the identification and
exploitation of community structure within a Collaborative Tagging System. It will
be shown that existing online applications can benefit from incorporating the results
of community detection and that new intelligent services can be developed on top of
the community analysis results. Moreover, an extension of a popular community
detection method [ 4 ] will be presented. The method is applied on tag networks in
order to discover sets of tags that are consistently used together by users of a Colla-
borative Tagging System and correspond to emerging topics of social interest. An
evaluation study on three real-world Collaborative Tagging Systems will be pre-
sented in order to demonstrate the value of the derived community analysis results.
The rest of the chapter has the following structure. Section 5.2 presents back-
ground information on the chapter topic. Section 5.3 presents an efficient commu-
nity detection scheme that addresses the specific characteristics of tag networks.
Furthermore, the section discusses how to evaluate the detected community

1

http://delicious.com

2

http://www.flickr.com
3 http://bibsonomy.org