Cognitive analytics for graphical legacy documents

Abstract

A cognitive cartography system receives a set of legacy documents, such as maps, in the form of physical hardcopy or as simple graphic images. The system, using rules derived from prior training and experience, revises the documents to resolve formal inconsistences like differences in resolution, orientation, or scale. The system assembles the adjusted documents into a seamless composite document represented as a computerized model. Applying learned rules and logic to contextual information received from extrinsic sources, the system infers semantic meaning from features represented by the composite, such as geographical features of a map. These inferences allow the system to derive new knowledge about the represented features, which is added to the model. When additional documents or contextual information are received, the system further refines the model by repeating this procedure. When the model has been sufficiently refined, the system makes the knowledge available to downstream systems.