To speed up the process of sorting letter mail, the U.S. Postal Service uses optical character recognition (OCR) machines on machine-printed mail. Even machine-printed mail can have missing words or characters in the address block or contain erroneous words or characters, resulting in rejection by the OCR as unreadable mail. The current procedure for dealing with such cases of patron errors is to subject the rejected mail to manual sorting. Missing character zip codes alone account for 61 percent of address line sets classified as unrecognized addresses. Manual sorting is tedious if left to unaided human perception alone. This paper describes a logic-based framework for address interpretation and rectification of machine-printed address mail rejected by the OCR for various reasons. A logic-based format was used consistently through the model, and implemented in Prolog. Preliminary testing of this early version of the model on 50 randomly selected samples of machine-printed mail rejected by the OCR at the Columbia general mail facility showed that 27 (54%) out of the 50 samples were rectified successfully. A classification of the types of errors in the study sample is also presented.