In the recent years, the amount of available textual documents has drastically increased, specially thanks to the electronic communication means like the WEB. Hence, technics allowing to structure, classify and recognize parts of this huge data base will allow a better use the available informations. In this aim, Logical Induction is a powerful tool that helps discovering patterns (and their structure) in textual data. In this paper, we present two applications of the Grammatical Inference (GI) : one applied on the Pattern Recognition aspects of a Document Recognition system; the second reports the use of the GI in a machine learning system applied to a textual data base. We specify and use the Grammatical Inference expressed as an instance of the Constraint Satisfaction Problem and then produce a Constraint Logic Program (CLP) to instantiate automata in a (language inclusion) lattice.