First International Conference on Spoken Language Processing (ICSLP 90)

Kobe, Japan
November 18-22, 1990

A Study on Various Prosody Styles in Japanese Speech Synthesizable With the Text-To-Speech System

Ryunen Teranishi

Kyushu Institute of Design, Shiobaru, Minami-ku, Fukuoka, Japan

This paper describes an original text-to-speech system developed in my laboratory, for the purpose of research study in the field of speech synthesis by rule, especially of synthesizing various prosody in Japanese. After many trials of the synthesis experiment with this system, it is found that most of prosody styles and the variations which appear in the text reading mode speech are synthesized, but prosody styles in another speech mode, viz. conversation mode are hardly synthesized, and only acted conversation style can be simulated. According to such results, the author discusses the synthesis possibility of various prosody styles, and he remarks that the speech prosody style of the text-reading mode (including the acted conversation style) is easier to be simulate with a simpler rule system, because the speech to be simulated is already regularized when the original speech turned to the literal text and turned again from letters to speech.

Full Paper

Bibliographic reference.  Teranishi, Ryunen (1990): "A study on various prosody styles in Japanese speech synthesizable with the text-to-speech system", In ICSLP-1990, 785-788.