The Method:
===========
The texts have been represented as character n-gram graphs of various
n-gram ranks. They have been clustered according to a heuristic where
a parent is inserted for every pair of vertices (documents) that are
most similar one to another, but have lower similarity that the
average similarity between all pairs of documents plus the standard
deviation of the latter. The clustering is performed on weight
(i.e. similarity) sorted pairs (edges) ascendingly. 

The Group:
==========
- George Giannakopoulos
- Ilias Zavitsanos
Both affiliated to National Center for Scientific Research Demokritos,
Software and Knowledge Engineering Laboratory, Greece and University
of Aegean, Department Of Information and Communication Systems
Engineering, Greece