A New Postprocessing Algorithm Based on Statistical Language Model and Lexicon

Abstract

This paper presents a new kind of postprocessing algorithm for a Chinese OCR. In this algorithm, a statistical language model and a word dictionary are used as linguistic information to improve the performance of the OCR. The linguistic information is used to form the sentence candidates, to prune the unreasonable character combination and to evaluate the ultimate sentence candidates. Some experiments have been conducted to verify the algorithm. The experimental results show that the algorithm significantly improves the recognition rate.