利用双向SOFM网络模拟汉字认知过程的研究  

Research of simulating recognition of Chinese characters by Bidirectional SOFM

作　　者：艾轶博[1] 穆志纯[1] 陈静[1] 

机构地区：[1]北京科技大学信息工程学院,北京100083

出　　处：《计算机应用》2006年第12期2971-2973,2976,共4页journal of Computer Applications

摘　　要：在汉字的认知过程中有“字优效应”和“字劣效应”,前者认为在汉字认知过程中整字信息优于部件或笔画信息,后者反之。以自组织特征映射算法为理论基础,提出了一种双向自组织特征映射(SOFM)网络,利用自组织网络实现根据汉字和部件多维表征的聚类,并建立两层网络之间的连接关系,通过双向测试,得到不同构型汉字所具有的字优效应和字劣效应,从新的角度实现了SOFM的应用。研究结果对于汉字教学方法有一定的参考价值。Word superiority effect and word inferiority effect are invovled in recognition of Chinese characters. The former one means that the recognition of holistie character unit is better than that of radical unit and the latter contrarily. In this paper, bidirectional SOFM was proposed based on Self-Organization Feature Map Mgorithm, accomplishing clustering of Chinese characters and their radical unit, establishing the connection between the two nets. Through bidirectional test, diversity constructional Chinese characters have their own effects, and SOFM is used in a new perspective. The result is helpful to Chinese characters teaching.

关 键 词：自组织特征映射 汉字认知 聚类 字优效应 字劣效应 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]