A Graphic Matching Process for Searching and Retrieving Information in Digital Libraries of Manuscripts

This paper outlines ICRPad, a pattern recognition system based on a graphic matching algorithm, which works on images by shape contour recognition, without requiring any segmentation process. The algorithm starts the process from a region of interest (ROI) selected in the image, using it as a shape model and looking for similar patterns in one or many target images. The process was developed and tested with the aim of proposing a new approach for searching and retrieving information in digital libraries. This approach is based on the application of data science, the fourth paradigm of knowledge development in the scientific field, that is at the basis of science informatics, to studies in data humanities. Following this approach, the algorithm is applied to find new research hypotheses through the discovery of patterns directly inferred from large digital libraries.