Social Network Analysis in Community-Built Databases - Community-Built Databases: Research and Development - page 63

applications, it is possible to provide additional information about key properties
interesting to the user, which can be used to filter unsuitable concepts during the
lattice construction [ 3 ]. In some applications, it is also possible to select one
particular concept and navigate through its neighborhood. These approaches allow
us to manage a larger scale of data but cannot provide the whole picture of a lattice.
A lot of social networks can be seen as object-attribute data or simply a matrix
(binary and fuzzy). They can be processed using matrix factorization methods,
which have been proven to be useful in many data mining applications dealing with
large-scale problems. Our aim is to allow the processing of a larger amount of data
and our approximation approach is compatible with the two mentioned in the
previous paragraph.
Clearly, some bits of information have to be neglected, but we want to know how
close or far from the original result we are. One way of doing this would be to
directly compare the results from the original and reduced datasets. Egghe and
Rousseau [ 14 ] introduce a modification of the classic Lorenz curve to describe the
dissimilarity between presence-absence data.
3.1.8 Related Work
Previously, the social network approach was applied in connection with CBD,
mostly with the analysis of Wikipedia. In this section, we refer to some of these
publications.
The SNA metrics computed on Wikipedia ([ 40 ]) is used mainly in the area of
content quality evaluation [ 23 ]. Suh et al. [ 36 ] identify user conflicts based on
mutual action reverts. Crandall et al. [ 10 ] analyze user attributes and their behavior
with respect to their friends. A Wikipedia article description using edit network is
contained in [ 5 ]. An older, well-known approach of the history of visualization
using the so-called history flow can be found in [ 39 ]. Finding experts both from the
Wikipedia content and among users is in [ 13 ].
An interesting prototype of CBD enhancements using social interaction and
online communities can be found in [ 2 ]. An automatic recommendation of work
that has to be done according to the user profile is presented in [ 9 ]. Said et al. [ 34 ]
tries to explain several CBD properties using the social properties of its users.
We have selected the following experiments to illustrate both the usefulness and
the bottlenecks of the social point of view in CBD. We use FCA for structure
analysis and matrix reduction methods to handle the complexity.
3.2 Tools and Techniques
Before we present our experiments illustrating the usefulness of the social perspec-
tive on CBD, we have to recall some basic notions of tools and techniques we
will use.