Intelligent OCR editor

This paper introduces a AI-based OCR post-processing technique, implemented as an Intelligent OCR Editor (IOCRED), which could enable the automation of OCR post-processing procedure and, therefore, could result in the increase of throughput, the decreases of error rate and the reduction of cost per page of an OCR system. An IOCRED system consists of a number of commercially available but different OCR systems performing text conversion on the same page simultaneously; a comparator detecting errors in the conversion results of OCRs; and, an intelligent error correction system using AI techniques such as expert systems, neural networks and fuzzy logic, to correct the detected errors. The IOCRED system is based on the premise that different OCR algorithms have distinct error characteristics. Such distinctions can be utilized by a cognitive device to detect and correct the errors in the conversion results of OCRs. To prove the concept, a statistical analysis of the error characteristics of OCR systems was conducted. Three popular commercial OCR systems were chosen for the study. The results showed that these OCR systems have distinct error characteristics and it is possible to achieve a high accuracy OCR conversion utilizing these differences. A simulation system used to examine the performance of the proposed IOCRED system was developed. The results of the simulations showed that utilizing the IOCRED system to achieve a high throughput, low error rate and low cost OCR conversion can be expected