ExpressiveSpeech / README_zh.md
Linz13
update
8348ba6

ExpressiveSpeech 数据集

项目网页

English Version

关于数据集

ExpressiveSpeech 是一个高质量、富有表现力双语(中英)语音数据集,旨在解决现有对话数据集中普遍存在的语音表达力不足的问题。

该数据集精心整理自五个著名的开源情感对话数据集,分别是:Expresso、NCSSD、M3ED、MultiDialog 和 IEMOCAP。 通过严格的处理和筛选流程,ExpressiveSpeech 确保了每一段语音都在音频质量和情感丰富度上达到高标准。 此数据集专为需要高保真度、能引起情感共鸣音频的任务而设计,例如富有表现力的语音转语音(S2S)、文本转语音(TTS)、语音转换以及语音情感识别等领域。

主要特点

  • 高表现力: 平均表现力得分高达 80.2,显著优于其原始来源数据集。
  • 双语内容: 包含均衡的中英文语音内容,语言比例接近 1:1
  • 数据规模可观: 包含约 14,000 条语音片段,总时长达 51 小时。
  • 丰富的元数据: 为每条语音提供由 ASR 生成的文本转录、表现力得分以及来源信息。

数据集统计

指标 数值
总语音条数 ~14,000
总时长 ~51 小时
语言 中文, 英文
语言比例 (中:英) 约 1:1
采样率 16kHz
平均表现力得分 (DeEAR) 80.2

我们的表现力评分工具:DeEAR

本数据集的高表现力是使用我们的筛选工具 DeEAR 实现的。如果您需要自己构建更大批量的高表现力数据,欢迎使用此工具。您可以在我们的 GitHub 上找到它。

数据格式

数据集的组织结构如下:

ExpressiveSpeech/
├── audio/
│   ├── M3ED
│   │    ├── audio_00001.wav
│   │    └── ...
│   ├── NCSSD
│   ├── IEMOCAP
│   ├── MultiDialog
│   └── Expresso
└── metadata.jsonl
  • metadata.jsonl: 一个 jsonl 文件,其中包含每条语音的详细信息。元数据包括:
    • audio-path: 音频文件的相对路径。
    • value: 由 ASR 生成的文本转录。
    • emotion: 来自原数据集的情感标签。
    • expressiveness_scores: 来自 DeEAR 模型的表现力得分。

JSONL 文件示例

每行 JSONL 包含一个 conversations 字段,其中存放语音轮次数组。 示例:

{"conversations": [{"No": 9, "from": "user", "value": "Yeah.", "emotion": "happy", "length": 2.027, "score_arousal": 0.9931480884552002, "score_prosody": 0.6800634264945984, "score_nature": 0.9687601923942566, "score_expressive": 0.9892677664756775, "audio-path": "audios/Expresso/splitted_conversation/ex04-ex01/laughing/ex04-ex01_laughing_001/009_speaker1_53s_55s.wav"}, {"No": 10, "from": "assistant", "value": "What was the reason, what was the, why couldn't I get there, ah I forget.", "emotion": "happy", "length": 3.753, "score_arousal": 0.9555678963661194, "score_prosody": 0.6498672962188721, "score_nature": 1.030701756477356, "score_expressive": 0.9965837001800537, "audio-path": "audios/Expresso/splitted_conversation/ex04-ex01/laughing/ex04-ex01_laughing_001/010_speaker2_55s_59s.wav"}]}
{"conversations": [{"No": 10, "from": "user", "value": "What was the reason, what was the, why couldn't I get there, ah I forget.", "emotion": "happy", "length": 3.753, "score_arousal": 0.9555678963661194, "score_prosody": 0.6498672962188721, "score_nature": 1.030701756477356, "score_expressive": 0.9965837001800537, "audio-path": "audios/Expresso/splitted_conversation/ex04-ex01/laughing/ex04-ex01_laughing_001/010_speaker2_55s_59s.wav"}, {"No": 11, "from": "assistant", "value": "Because genie really had to go and and to the bathroom and she couldn't find a place to do it and so she when they put the tent on it it was it was a bad mess and they shouldn't have done that.", "emotion": "happy", "length": 10.649, "score_arousal": 0.976757287979126, "score_prosody": 0.7951533794403076, "score_nature": 0.9789049625396729, "score_expressive": 0.919080913066864, "audio-path": "audios/Expresso/splitted_conversation/ex04-ex01/laughing/ex04-ex01_laughing_001/011_speaker1_58s_69s.wav"}]}

:部分源数据集使用了 VAD(语音活动检测),可能会将一句话切分为多个片段。 为保持对话完整性,我们通过特定规则将这些片段合并回完整语句。

授权协议

为遵循其源数据集的非商业性使用限制,ExpressiveSpeech 数据集在 知识共享署名-非商业性使用-相同方式共享 4.0 国际 (CC BY-NC-SA 4.0) 许可协议下发布。

您可以在此处查看完整的许可协议。

如何引用

如果您在研究中使用了本数据集,请引用我们的论文:

@article{lin2025decoding,
  title={Decoding the Ear: A Framework for Objectifying Expressiveness from Human Preference Through Efficient Alignment},
  author={Lin, Zhiyu and Yang, Jingwen and Zhao, Jiale and Liu, Meng and Li, Sunzhu and Wang, Benyou},
  journal={arXiv preprint arXiv:2510.20513},
  year={2025}
}