Journal of the Phonetic Society of Japan
Online ISSN : 2189-5961
Print ISSN : 1342-8675
Feature Articles: Data-Driven Phonetic Analysis Using Large-Scale Corpora
Prosodic Speaking Styles Extracted from the X-JToBI Annotation of the Corpus of Spontaneous Japanese(<Feature Articles>Data-Driven Phonetic Analysis Using Large-Scale Corpora)
Kikuo MAEKAWA
Author information
JOURNAL FREE ACCESS

2014 Volume 18 Issue 1 Pages 70-82

Details
Abstract
Prosodic features are often neglected in the stylistic analyses of spoken language. In this paper prosodic speaking style were automatically extracted from the label frequency data of the X-JToBI prosodic annotation as applied to the CSJ-Core. Principal component analysis of the data revealed three basic prosodic speaking styles. The first style is characterized by the use of boundary tone L% and the BI 2 (ordinary accentual phrase boundary). The second style is characterized by the combination of H% boundary tone and two BIs, 2 + b and 2 + bp (accentual phrase with boundary pitch movement, but without pitch resetting). The third style is characterized by the combination of HL% and BI 3 (intonation phrase boundary with pitch resetting). It also turned out that the second and third speaking styles showed almost opposite behavior when they are correlated with the Impression Rating Score data of the CSJ-Core. The second style tended to be evaluated as polite and punctual, while the third style tented to be evaluated as casual and relaxed. Lastly, it was shown that the impression of the utterances having the PNLP variant of the HL% boundary tone (that has its peak on the penult mora unlike the the ordinary variant of the HL% that has the peak on the last mora) were more similar to the impression of the second style than that of the third style.
Content from these authors
© 2014 The Phonetic Society of Japan
Previous article Next article
feedback
Top