A fuzzy method for incorporating the contextual constraints into a text recognition system is presented here. The method takes as input all the internal result that an Isolated Character Classifier (ICC) computes for an input letter, instead of an unique output character. The internal result is handled here as a fuzzy set which is then processed by a Deformed System. Such a Deformed System represents a dictionary of legal words, and it is actually a Finite Automaton which has been modified for accepting as input no single symbols but fuzzy sets. Several tests have been carried out in a Text Recognition System and the obtained results show the suitability of the method.