It is well known that the convergence of a grammatical inference method is strongly conditioned by the training data set. Structural completeness is a desired property seldom achieved in real data. The question that naturally arises in these types of problems is: how far is the training data to achieve structural completeness and what is the minimal sample size to use when there is no a priori knowledge about the structure of the data. In this paper we propose a simple methodology to give some

more »

... nsight into the later problem. It basically consists of a bootstrapping technique supported on grammars inferred from the existing data. An example of the application of this methodology in the context of automatic sleep analysis is used to illustrate the method. * This work was partially supported by the project PECS/C/SAU/212/95 and BRAXIS 2/2.1/T~T/l~80/95.