A system for automated data entry from forms

OCR technology cannot on its own, meet the demands of automated data entry. Two types of postprocessing are generally required. First, context must be used to improve the recognition and accept some fields automatically. Second, operator intervention is needed to verify some results or enter fields that are too difficult to recognize. In this paper we describe a novel approach to integrate both types of post-processing into a single system. Context analysis and the interactive functions are invoked alternatively several times as needed. Each step improves the results until all fields are certified. The paper also discusses a framework for evaluating the efficiency of the method and provides some experimental results