Tohoku University Technology: Large-scale Emotional Speech Corpus 'JTES': S20-017
Japanese speech database available for research and development in AI and voice recognition.
Many voice dialogue systems that utilize AI, including chatbots, only handle linguistic information of the spoken content, which can lead to challenges such as conversations not being established or feeling unnatural depending on the dialogue content. JTES (Japanese Twitter-based Emotional Speech) is a general-purpose emotional voice corpus designed for use in voice dialogue systems that realize "emotion recognition," which estimates emotions from the tone of the input voice; "emotional voice recognition," which recognizes speech in emotional voices; and "emotional voice synthesis," which synthesizes speech imbued with emotion. Specifically, it contains 20,000 utterances (23.5 hours) spoken by 100 general speakers (50 males and 50 females), each producing 50 sentences for four emotions (joy, anger, sadness, and neutrality). JTES can be utilized for research and development of high-precision emotion recognition and expressive synthetic voices.
- Company：Tohoku Techno Arch Co., Ltd.
- Price：Other
