Emotive Narrative Storytelling (EMNS) corpus
收藏资源简介:
EMNS数据集是由伯恩茅斯大学的数字娱乐中心创建,旨在提升交互式叙事驱动系统中的语音表达性和情感质量。该数据集包含2.3小时的录音,由一名女性演讲者提供,涵盖八种表演情感状态,情感分布均匀,具有0.68%的变异度。数据集还包括表达性水平和自然语言描述,附带单词强调标签。EMNS数据集通过精心选择Mozilla的Common Voice数据集中的句子,确保了语音的多样性和未来扩展的可能性。此外,数据集的创建过程中使用了Montreal Forced Aligner进行语音和文本的对齐,以提高语音合成的自然度和表达性。该数据集适用于语音合成研究,特别是在虚拟助手、交互式故事讲述和娱乐平台等领域,旨在通过合成语音有效传达情感,创造更吸引人的用户体验。
The EMNS dataset was created by the Digital Entertainment Centre at Bournemouth University, aiming to enhance the vocal expressiveness and emotional quality of interactive narrative-driven systems. This dataset contains 2.3 hours of recorded speech provided by a female speaker, covering eight acted emotional states with uniform emotional distribution and a variability of 0.68%. The dataset also includes expressiveness levels and natural language descriptions, accompanied by word-level emphasis tags. The EMNS dataset ensures vocal diversity and potential for future expansion by carefully selecting sentences from Mozilla's Common Voice dataset. Additionally, the Montreal Forced Aligner was utilized during the dataset creation process to align speech and text, thereby improving the naturalness and expressiveness of speech synthesis. This dataset is applicable to speech synthesis research, particularly in fields such as virtual assistants, interactive storytelling, and entertainment platforms, with the goal of effectively conveying emotions through synthesized speech to create more engaging user experiences.

- 1EMNS /Imz/ Corpus: An emotive single-speaker dataset for narrative storytelling in games, television and graphic novels数字娱乐中心,伯恩茅斯大学,英国 · 2023年



