Contextual Postprocessing Expert System for English Sentence Reading Machines

摘要

Knowledge-based programming techniques are used in an expert system to reduce uncertainity for optical character recognition by combing evidence from several diverse knowledge sources cooperating in an hierarchical configuration. The postprocessor development focused on a system for generic text input that is not constrainted to a fixed vocabulary or particular subject domain. The key element to the system's effectiveness is a spell checking algorithm that is not operationally bounded by an enumerated lexicon or biased by statistical sampling. The postprocessor system is also design to interface almost any type of OCR front-end methodology. Keywords: theses; image processing.