Figure 9 shows a typical tree produced using SCENT with 101 untyped vectors of the
Zoo data set. Each of the data vectors is clearly categorised at both super and subordinate
levels. The tree structure produced using each data set was labeled in two
ways. Firstly, the data vectors represented by each leaf node of the tree were labeled
with the associated animal name tag. Secondly, the semantic type label of each instance
was examined and the leaf node and super-ordinate clusters were classified according to
each of the 7 labeled groups. Each super-ordinate class is further classified at each
subsequent level of the tree. All instances of mammals are grouped together, in the
leftmost clusters of the tree, and are subdivided into large-predatory, large-nonpredatory,
small, aquatic-legged, aquatic-finned. The middle cluster categorises fish
and birds. The birds are classified in two subtrees, as predatory or non-predatory. The
right hand cluster represents a variety of vectors from the less well represented classes
of animal. It can therefore be seen that the emergent clusters produced by SCENT
often correspond with natural taxonomic groups.