Using intuitionistic fuzzy sets in text categorization

Abstract

We address some crucial problem associated with text categorization, a local feature selection. It seems that intuitionistic fuzzy sets can be an effective and efficient tool making it possible to assess each term (from a feature set for each category) from a point of view of both its indicative and non-indicative ability. It is important especially for high dimensional problems to improve text filtering via a confident rejection of non-relevant documents. Moreover, we indicate that intuitionistic fuzzy sets are a good tool for the classification of imbalanced and overlapping classes, a commonly encountered case in text categorization. © 2008 Springer-Verlag Berlin Heidelberg.