It depends upon the probabilistic relationship between
different categories. If it is considered as a graph then there is
a parent node, which is connected to the child nodes, other
than this no other connection is possible. It is easy and
efficient to implement, because it assumes that all the words
of the document are independent to each other.