In this paper, we consider the Pattern Recognition applied to paper documents based on the grammatical inference (GI) for classes of structured documents like summaries, dictionaries, bibliographic data basis, encyclopaedias and so on. In this task, the inference engine takes as input a set of individual examples of these documents and outputs a set of rules that recognise similar documents. We place GI in an algebraic framework in which rewrite rules will define the process of generalisation. The implementation algorithm discussed here is used in a current document handling project in which paper documents are typographically tagged and then recognised. One of the current applications in this project is to extract the physical and the logical structures of a given set of paper documents and then reorganise them in a machine readable form like HTML code.