| This research explores a hybrid knowledge-based system for detecting defects in digital documents, especially defects that depend on the style of the document designer. The hybrid system is composed of a rule-based engine in charge of the document segmentation and understanding, and a case-based reasoning engine in charge of recognizing the defects. The system performs with an 81.2% of accuracy in the recognition of defects. This thesis includes the architecture design for the hybrid system, the characterization of defects, the methodology for document segmentation and the extraction of the document implicit knowledge, and finally the experimental results of defect detection. |
