摘要
总结了英文光学字符识别 (OCR)系统中的常见错误类型 ,论证了使用拼写检查技术发现错误的可能性 .以最相似单词替换OCR的错误输出 ,从而达到改正错误的目的 .利用编辑距离实现词汇相似性评估 ,并用动态规划技术优化了编辑距离的计算 .试验结果证明该方法可以使OCR系统的单词识别率提高 6 2 5 % .
Putting up the type of the common errors from the system English OCR,it has proved it feasible to use a spelling check technology to find its errors.Using the most similar words to replace OCR's error output,to correct such errors,using edit distance to evaluate the similaring of words,and using dynamic planning technology to optimize the calculation of edit distance,the recognition rate of words from OCR can be improved by 6 25%.
出处
《鞍山钢铁学院学报》
2002年第3期192-196,共5页
Journal of Anshan Institute of Iron and Steel Technology
