A fuzzy-rough approach to the representation of linguistic hedges

We present a new representation for linguistic hedges using a framework of fuzzy rough sets. In traditional fuzzy-set theoretical representations, properties of objects such as old and experienced, axe represented by a fuzzy set P, while linguistic hedges (i.e. expressions like very, more or less, rather) are modelled by means of some transformations applied to P. In contrast to these approaches, we propose a representation which allows us to express the meaning of a statement like "x is very P" also relative to mutual resemblances between objects in the domain of discourse. This allows for adequate context-dependent characteristics of objects. Technically, this is achieved by using fuzzy rough approximators with respect to fuzzy resemblance relations representing mutual resemblances between objects. We show that this framework allows for flexible representation of some linguistic terms