摘要
韵律规则主要包括基频和音长变化规律，目前的韵律规则大多是通过人工得到的一些定性描述．为了学习汉语两字词的韵律规则，文中将韵律规则看做是孤立音节语音和其在词中发音两者之间的映射关系，从而使用神经网络来学习这种映射．文中将介绍这种方法的神经网络结构并给出实验结果．
Prosodic model includes pitch model and duration model,and these models being used are described qualitatively.In order to learn the prosodic patterns from Chinese two word phrases,the patterns are regarded as the mapping between the isolate syllable and the same one when it is spoken in phrases,and the mapping can be learned by using neural network.In the paper here,the architecture of learning neural networks is introduced,and the results are also given.
出处
《计算机研究与发展》
EI
CSCD
北大核心
1999年第6期664-667,共4页
Journal of Computer Research and Development
基金
"中国科学院百人计划"项目基金
