An Intelligent System for Conflict Resolution in Handwritten Address Recognition

Background: The basic recognition engine of a handwritten address interpretation system, for use in postal sorting automation, is an OCR algorithm that recognises a numeric or alphanumeric string, such as a postcode, and matches it against a set of valid postal delivery points. However, the OCR system is highly vulnerable to errors due to the uncertainty that arises when the imperfect OCR result of the alphanumerics are combined. These errors result in wrongly sorted mail which costs postal organizations throughout the world a significant amount of money to resolve after the error is detected by the postal delivery man and also delays the delivery of the wrongly-sorted item. A generic expert system model that forms part of a handwritten address interpretation system for conflict resolution in unconstrained handwritten address recognition is presented in this paper, to reduce this error. Method: The proposed expert system resolves the conflicts and reduces the error rates by fusing a holistic pattern recognition method with expert knowledge based on a posterior information. The system was evaluated using 1,071 handwritten Singapore addresses. Results: Experimental results show that the expert system achieved a significant reduction in error rates. Performance was improved from 71.2% correctly sorted, 4.8% reject (cannot sort) and 24.0% error (wrongly sorted) rates using OCR only to 63.7% correctly sorted, 35.7% reject (cannot sort) and 0.6% error (wrongly sorted) rates using the proposed approach. Conclusions: The error rate (proportion of wrongly sorted mail) can be significantly reduced using this method and significantly alleviate the need to carry out expensive resorting.