JVNV
收藏资源简介:
JVNV是一个包含日语情感语音的数据集,由东京大学信息科学与技术研究生院创建。该数据集包含514条语音数据,涵盖六种基本情绪,由四名专业演讲者录制,每条语音至少包含一个非语言声音(NVs)。JVNV数据集通过大型语言模型自动生成情感脚本,确保了语音数据的情感表达和语音覆盖的平衡。该数据集适用于语音情感识别和情感语音合成等任务,旨在解决现有情感语音数据集中情感脚本和非语言声音表达不足的问题。
JVNV is a dataset of Japanese emotional speech developed by the Graduate School of Information Science and Technology at The University of Tokyo. This dataset comprises 514 speech samples covering six basic emotions, recorded by four professional speakers, with each sample containing at least one non-verbal sound (NVs). The JVNV dataset automatically generates emotional scripts through Large Language Models, ensuring a balanced distribution of emotional expressions and speech coverage across the collected speech data. This dataset is suitable for tasks including speech emotion recognition and emotional speech synthesis, and it aims to address the shortage of adequate emotional scripts and non-verbal sound expressions in existing emotional speech datasets.




