The comparison between Thai and Japanese temporal control characteristics using segmental duration models

Abstract

This paper compares the temporal control characteristics between Thai and Japanese read speech data using segmental duration models. The same and the different control characteristics have been observed from phone level to sentence level. The language-dependent and language-independent control factors have also been observed. In phone and neighboring phone level, different characteristics are found. Japanese vowel durations are mainly compensated by only adjacent preceding and following phones, which results from mora timing. Unlike Japanese, Thai vowel durations are affected by two succeding phones. It can be guessed that the differences come from syllabic structures. In word level, most content words tend to have longer phone durations while function words have shorter ones. In phrase level, both languages express duration lengtening of syllable/mora at the phrase initial and final. For language-specific factors, Thai tones express small alteration on phone duration. The comparisions explore the duration characteristics of the languages and give more understanding to be used in speech synthesis and second-language learning research. [Work supported in part by Waseda Univ. RISE research project of ‘‘Analysis and modeling of human mechanism in speech and language processing’’ and Grant-in-Aid for Scientific Research A-2, No. 16200016 of JSPS.]