Community Detection in Collaborative Tagging Systems - Community-Built Databases: Research and Development - page 129

presents several usage contexts discovered for the tags “networks,” “bridge,” and
“bean” taken from the datasets BIBSONOMY-200K, FLICKR-1M, and DELICIOUS-
7M, respectively. 11 These contexts correspond to meaningful concepts/objects
connoted by the tags, for example, in the case of “bean,” there are tag communi-
ties corresponding to bean as a plant seed, coffee bean, java bean, etc. In the
future, we plan to investigate methods of accurately mining such tag contexts in an
automatic way.

5.5 Conclusions

This chapter provided an overview of community detection in Collaborative
Tagging Systems. It discussed a range of works in this area that deal with the
problem of identifying groups of objects (communities) in a tagging system
that are highly related to each other. The chapter structured the related work
discussion along three dimensions: type of community (user, resource, tag or
composite), community detection method, and application context. In addition, a
parameter-free extension of an existing community detection method was pre-
sented that is catered for the particular characteristics of tag networks. Further-
more, a series of means for evaluating the derived tag community structure were
discussed.
Finally, a comparative evaluation study was conducted involving the proposed
technique and an established modularity maximization scheme. Three datasets
coming from real-world tagging systems were employed in the study to ensure
the reliability of the findings. The proposed approach was demonstrated to produce
tag communities that are more precise and more useful for tag recommendation
than the ones produced by the modularity maximization scheme. In addition, the
presented method was successfully applied to discover multiple contexts of usage
for several hub tags.

Acknowledgments This work was supported by the WeKnowIt and GLOCAL projects, partially
funded by the European Commission, under contract numbers FP7-215453 and FP7-248984
respectively.

11

In fact, these hub tags were adjacent to many more tag communities than the ones presented. Out
of those some were very generic, some were similar to the ones presented, and for some we could
not establish a profound sense. For the sake of brevity, we manually picked some of the prominent
contexts to present.