Abstract
We study approximations of regular languages bymembers of a given variety ℒ of regular languages. These are upper or lower approximations in the sense of Pawlak's rough set theory with respect to congruences belonging to the variety of congruences corresponding to ℒ. In particular, we consider the closest upper and lower approximations in ℒ. In so-called principal varieties these always exist, and we present algorithms for finding them, but for other varieties the situation is more complex. Although we consider just Eilenberg’s +-varieties, the general ideas apply also to other types of varieties of languages. Our work may also be viewed as an approach to the characterizable inference problem in which a language of a certain kind is to be inferred from a given sample.