词类-音节数序列与语句韵律结构关联及韵律结构预测

发布时间：2018-02-11 20:36

本文关键词： 词类-音节数序列文语转换韵律结构预测　出处：《语言科学》2011年06期 　论文类型：期刊论文

【摘要】：提高汉语合成语音自然度和可懂度的最有效、最关键途径在于更准确地预测语句的韵律结构。目前文语转换系统预测语句韵律结构所能依据的内容通常只有语句的词类-音节数序列,文章主要讨论从词类-音节数序列到语句韵律结构预测间的关联,并据此对如何提高韵律结构预测准确率提出了一些建议:词类-音节数序列中包含的"韵律上最可能共存与最不可能共存的组合"以及"能够帮助有效预测停延段信息的虚词"可以提高预测准确率;对序列中某些片段语法结构的确定,可以提高预测准确率;最后,根据韵律分布特征细分词类,也可以提高预测准确率。
[Abstract]:The most effective way to improve the naturalness and intelligibility of Chinese synthetic speech, The most important way is to predict the prosodic structure of sentences more accurately. At present, the content on which the prosodic structure can be predicted by the text-to-speech conversion system is usually only based on the word class and syllable sequence of the sentence. This paper mainly discusses the correlation between speech class and syllable number sequence and sentence prosodic structure prediction. Some suggestions on how to improve the accuracy of prosodic structure prediction are put forward: the combination of the most likely and the most unlikely coexistence in prosody and the combination of the most likely and the most unlikely coexistence in part of speech and syllable sequence, and the combination that can help to effectively predict stoppage. The function word "of paragraph information" can improve the accuracy of prediction; The accuracy of prediction can be improved by determining the grammatical structure of some segments in the sequence. Finally, the prediction accuracy can be improved by subdividing the parts of speech according to the prosodic distribution features.
【作者单位】：西安外国语大学汉学院;
【基金】：教育部人文社科一般项目“现代汉语中韵律与语法的作用关系研究”(编号:10YJC740079) 陕西省教育厅项目“服务于对外汉语教师培养的汉语教学语法研究”(编号:2010JK249)
【分类号】：H116.3

【参考文献】