Construction of knowledge-based next generation document

字符识别技术是面向21世纪的一项重要技术。在本研究中，我们提出了一个基于知识的文档识别系统。分割技术、识别技术和基于知识的后处理技术是该系统的三个主要部分。最后，将该系统应用于手写地址识别系统中，取得了很高的识别率。测试样本为1840份，由邮电部收集。识别率在90%以上，漏检率在2%以下。拒绝率在18%左右。

Character recognition technology is an important technology for the new 21th century. In this research, we proposed a knowledge-based document recognition system. Segmentation technique, recognition technique and knowledge-based post-processing technique are the three main parts of this system. As the result, we applied the proposed system to a handwritten address recognition system and obtained a very high recognition rate. The test sample are 1,840, which collected by the Ministry of Post and Telecommunications. The recognition rate is beyond 90%, and the missed rate is below 2%. The reject rate is around 18%.