Empirical learning methods for digitized document recognition: an integrated approach to inductive generalization

A hybrid method of using empirical and supervised learning to acquire knowledge expressed in the form of classification rules is applied to optically scanned documents with the aim of automatic recognition and storage. An expert system devoted to classification recognizes a document as belonging to a class by its layout and the logical structure of a generic printed page. Decision rules for document classification are inferred by inductive generalization. The learning methodology combines a data analysis technique for linearly classifying with a conceptual method for generating disjunctive cover for each class of document.<
>