JVNV: A Corpus of Japanese Emotional Speech with Verbal Content and Nonverbal Expressions

Citation Author(s):
Detai
Xin
The University of Tokyo
Junfeng
Jiang
The University of Tokyo
Shinnosuke
Takamichi
The University of Tokyo
Yuki
Saito
The University of Tokyo
Akiko
Aizawa
National Institute of Informatics
Hiroshi
Saruwatari
The University of Tokyo
Submitted by:
Detai Xin
Last updated:
Mon, 10/09/2023 - 13:11
DOI:
10.21227/r008-1d72
Data Format:
License:
0
0 ratings - Please login to submit your rating.

Abstract 

JVNV is a Japanese emotional speech corpus with verbal content and nonverbal vocalizations whose scripts are generated by a large-scale language model.

Existing emotional speech corpora lack not only proper emotional scripts but also nonverbal vocalizations (NVs) that are essential expressions in spoken language to express emotions.

We propose an automatic script generation method to produce emotional scripts by providing seed words with sentiment polarity and phrases of nonverbal vocalizations to ChatGPT using prompt engineering.

We select 514 scripts with balanced phoneme coverage from the generated candidate scripts with the assistance of emotion confidence scores and language fluency scores.

JVNV has better phoneme coverage and emotion recognizability than previous Japanese emotional speech corpora.

To our best knowledge, JVNV is the first speech corpus that generates scripts automatically using large language models.

Instructions: 

JVNV consists of Japanese emotional speech with both verbal content and nonverbal vocalizations (NVs) such as laughter, and sobbing that can express emotions.

Each utterance of JVNV expresses a certain emotion, and has at least one phrase of NV.

In addition to the audio and transcriptions, the duration of each NV phrase is also provided

Funding Agency: 
Japan Science and Technology Agency, Japan Society for the Promotion of Science
Grant Number: 
JPMJSP2108, JP23KJ0828

Comments

.

Submitted by HYEONJOO KANG on Thu, 05/23/2024 - 03:10