论文部分内容阅读
为了探究神经网络模型在儿童语言习得模拟研究中的应用,该文在可扩展的自组织网络模型(growing selforganizing map,GSOM)算法的基础上,模拟了婴幼儿习得标准德语部分元音和辅音音位范畴的过程。该研究将优化的网络扩展策略和“循环性强化和复习训练”学习算法与传统的GSOM算法进行了结合。模拟结果显示:“循环性强化和复习训练”算法可以有效地提高模型网络的学习质量;模型算法可以较好地习得元音音位和辅音发音方式的范畴,并构建相应的知识网络。建模研究的结果表明:在习得语言的过程中,通过对语音声学信息的加工,婴幼儿有能力习得元音音位和辅音发音方式的范畴,并构建元音音位在声学空间内的分布关系。
In order to explore the application of neural network model in the study of children language acquisition simulation, based on the scalable selforganizing map (GSOM) algorithm, this paper simulates the standard vowels and vowels Consonants phonemes categories of process. The study combines optimized network expansion strategies and the “Cyclic Enhancement and Review Training” learning algorithm with the traditional GSOM algorithm. The simulation results show that the algorithm of “cyclical reinforcement and review training” can effectively improve the learning quality of the model network. The model algorithm can better acquire the categories of vowel phonemes and consonant pronunciation, and construct the corresponding knowledge network . The results of the modeling study show that in the process of acquiring a language, infants and young children have the ability to acquire the categories of vowel phonemes and consonant pronunciation through the processing of speech and acoustic information, and construct the vowel phonemes in the acoustic space The distribution of relations.