节点文献
一种新的针对汉语TTS的韵律词自动切分方法
A New Automatic Segmentation Method of Prosodic Word for Chinese TTS
【Author】 Li Xiaoming Sun Jincheng(Chinese Academy of Sciences, Institute of Acoustics, Beijing 100080, China) (Department of Thermal Engineering, Tsinghua University, Beijing 100084, China)
【摘要】 以往的语音合成中,分词单位多为词典词,而这种分词单位所形成的韵律信息与TTS所需的韵律信息常不符合。本文针对TTS系统的需要制定了韵律词的标准。分别利用基于规则的方法和基于机器学习的方法对实际文本进行了韵律词级别的自动切分,讨论了韵律词自动切分的前景和难点。
【Abstract】 In former TTS system, the most popular segmentation unit is word from lexicon. But prosodic information based on this segmentation unit is not fit for TTS system. The article redefines the prosodic word for TTS system. The method based on regulars and the method based on machine learning are used to proceed automatic segmentation. The foreground and the difficulty of automatic segmentation of prosodic word are discussed.
- 【会议录名称】 第一届学生计算语言学研讨会论文集
- 【会议名称】第一届学生计算语言学研讨会
- 【会议时间】2002-08
- 【分类号】TN912.3
- 【主办单位】中国中文信息学会