An Efficient Post-processing Approach for Off-Line Handwritten Chinese Address Recognition

摘要

Language model is widely used in OCR post-processing. In this paper, based on language model, we propose a two-step method for post-processing of off-line handwritten Chinese address recognition. According to the characteristics of high level and low level addresses, two different strategies are used: for high level addresses, word-based language model is used with weighed matching search algorithm proposed; for low level addresses, ME model is implemented with similar character method. After post-processing, recognition accuracy rises from 41.60% to 78.19%, which means 62.65% errors reduction.