计算机模拟汉字字形认知过程的研究  

Computer simulation of the cognition of Chinese characters

作　　者：陈静[1] 穆志纯[1] 孙筱倩[1] 

机构地区：[1]北京科技大学信息工程学院,北京100083

出　　处：《智能系统学报》2008年第3期216-221,共6页CAAI Transactions on Intelligent Systems

基　　金：北京市教委重点学科共建基金资助项目(XK100080537);北京语言大学规划资助项目(04GH01)

摘　　要：对汉字的认知研究不仅是认知科学、也是计算机科学特别是人工智能领域中的一个研究热点.但是,目前汉字认知的计算机模拟研究还相对滞后,其在认知科学研究中的作用还无法和行为实验研究等同.从认知科学的角度出发,建立汉字字形表征库,构建模型,确定训练和测试方式等,对汉字字形认知过程(学习发展历程)中汉字聚类与部件拆分意识进行了计算机模拟,以便研究汉字字形学习中的某些认知规律.通过对模型的训练与测试,得到了输入汉字的聚类效果图、部件拆分情况,以及对模型进行生字测试的结果.得出的结果能够反映某些汉字认知的规律,所以模型在一定程度上模拟了汉字字形的认知过程.Research on the cognition of Chinese characters is a hotspot in both cognitive science and computer science, and is an especially lively field among those investigating artificial intelligence. In spite of this, research using computer simulations to analyze Chinese character cognition remains relatively backward, and its value in cognition studies has not been comparable with experimental research on behavior. In this paper, starting from the viewpoint of cognition science, a representative database of Chinese characters was set up, a cognitive model con- structed, and training and testing modes determined. Computer simulations were made of the clustering and split- ting of Chinese characters in the course of cognition, so that cognition rules for the perception of Chinese characters may be better understood. The model was based on a multi-layer self-organizing neural network. This training and testing method ensured that we knew how the Chinese characters were clustered and split during analysis so that the recognition of unknown words could be achieved. The research outcome suggests cognition rules for recognizing Chi- nese characters, implying that the proposed model does simulate the cognition process for Chinese characters.

关 键 词：认知科学 人工智能 汉字认知 计算机模拟 白组织模型 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]