Abstract:
This paper proposes to implement Bi-
Dimensional Probabilistic Model (BDPM). This
model is to find a bi-dimensional representation of
textual documents for the problem of text
categorization. The main idea is to consider the
importance of a word. In this model, terms are
seen as disjoint events, and terms and categories
are related to each other. Terms are measured
with their presence and expressiveness. The
presence and expressiveness of a term is defined
as the peculiarity of that term. BDPM is document
classification model which can represent a
document in two dimensional schemes. It can also
be used for both visualization and classification of
textual documents. To experiment the system,
Reuters-21578 newswire dataset is used.Thisbidimensional
model has the advantage: no explicit
need for feature selection, to reduce
dimensionality, since documents are represented
by only two dimensions.
