汉语作为第二语言的声调认知发展模拟

Simulating Mandarin tone acquisition by CSL learners who are native English speakers

该文提出了一个动态的生长型树型自组织特征映射网络,用以模拟英语母语者汉语声调的认知发展。该网络既克服了传统的Kohonen自组织特征映射网络的固定网络结构限制以及容量有限性,又克服了其他类似动态网络的较弱的拓扑映射特性,可以较好地模拟英语母语者汉语声调认知的发展。模拟结果跟实验结果呈现出非常好的一致性,既证明了行为实验中汉语声调的动态发展过程,也为汉语声调认知的机制研究提供了机理上的解释。

A growing tree-structured self-organizing feature map network was developed to simulate the acquisition of Mandarin tones by native English speakers.This network has a dynamic network construction and capacity that is better than the traditional Kohonen self-organizing feature map.This network also more effectively preserves the global topology mapping than early dynamic tree-like maps.Simulations of the acquisition of Mandarin tones are consistent with experimental results.These show the dynamic developmental process of Mandarin tones.The results show that self-organization is an important mechanism for tone emergence and that tone feature teaching is important for Mandarin tone-teaching strategies.