| Optical Character Recognition (OCR) is used to convert paper documents into electronic form. Unfortunately the technology is not perfect and the output can be erroneous. Conversion then is generally augmented by manual error detection and correction procedures which can be very costly.; One approach to minimizing cost is to apply an OCR post processing system that will reduce the amount of manual correction required. The post processor takes advantage of knowledge associated with a particular project.; In this thesis, we look into the feasibility of using integrity constraints to detect and correct errors in forms recognition. The general idea is to construct a database of form values that can be used to direct recognition and consequently, make automatic correction. |
