【摘要】 汉语是一种开放大字符集语言,汉字字形相似度计算是汉语信息处理的一项基础研究,对于汉字识别、计算机辅助的汉语文章校对和汉字教学都有重要作用.本文对现有汉字字形结构描述方法从图形相似角度进行了改进,并给出了一种基于结构描述的字形相似度计算算法,该方法计算相似度无需字形样本实例的学习训练,对于常用字和难于获取书写样本的生僻字的相似度计算,都具有很好的适应性,可满足不断扩大的汉字集合计算相似度的需要.实验表明,采用此法计算得到的GB2312中6763个汉字的相似字表,与人的认知结果吻合度很好,并应用于计算机辅助校对系统中的别字修改提示,显示出较好效果.更多还原

【Abstract】 Chinese is a language with a large open character set.The similarity calculation of Chinese character glyph is important in automatic recognizing,computer aided proofreading,and teaching of Chinese characters.The current description method of Chinese character glyph structure is improved from the perspective of graphical similarity and an algorithm for the calculation of glyph similarity of Chinese characters without training samples of character glyphs is proposed.This algorithm possesses a better adaptability for character similarity calculation of both common-used Chinese characters and rarely-used ones whose writing samples are difficult to obtain,so as to meet the demand for character similarity calculations on the continually expanding Chinese character set.Experiment indicates that the similar character lists of 6763 characters in GB2312 calculated by this algorithm have a high coincidence with human perception.The application of the similar character lists in a computer aided proofreading system improves the modification guide for users.更多还原