Linz13
commited on
Commit
·
f4dde24
1
Parent(s):
a953dcc
update
Browse files- README_zh.md +24 -4
README_zh.md
CHANGED
|
@@ -8,9 +8,9 @@
|
|
| 8 |
|
| 9 |
## 关于数据集
|
| 10 |
|
| 11 |
-
**ExpressiveSpeech**
|
| 12 |
|
| 13 |
-
|
| 14 |
|
| 15 |
## 主要特点
|
| 16 |
|
|
@@ -27,9 +27,15 @@
|
|
| 27 |
| 总时长 | ~51 小时 |
|
| 28 |
| 语言 | 中文, 英文 |
|
| 29 |
| 语言比例 (中:英) | 约 1:1 |
|
| 30 |
-
| 采样率 | 16kHz
|
| 31 |
| 平均表现力得分 (DeEAR) | 80.2 |
|
| 32 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
## 数据格式
|
| 34 |
|
| 35 |
数据集的组织结构如下:
|
|
@@ -48,11 +54,25 @@ ExpressiveSpeech/
|
|
| 48 |
```
|
| 49 |
|
| 50 |
- **`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:
|
| 51 |
-
- `
|
| 52 |
- `value`: 由 ASR 生成的文本转录。
|
| 53 |
- `emotion`: 来自原数据集的情感标签。
|
| 54 |
- `expressiveness_scores`: 来自 **DeEAR** 模型的表现力得分。
|
| 55 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 56 |
## 授权协议
|
| 57 |
|
| 58 |
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。
|
|
|
|
| 8 |
|
| 9 |
## 关于数据集
|
| 10 |
|
| 11 |
+
**ExpressiveSpeech** 是一个高质量、**富有表现力**的**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。
|
| 12 |
|
| 13 |
+
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。
|
| 14 |
|
| 15 |
## 主要特点
|
| 16 |
|
|
|
|
| 27 |
| 总时长 | ~51 小时 |
|
| 28 |
| 语言 | 中文, 英文 |
|
| 29 |
| 语言比例 (中:英) | 约 1:1 |
|
| 30 |
+
| 采样率 | 16kHz |
|
| 31 |
| 平均表现力得分 (DeEAR) | 80.2 |
|
| 32 |
|
| 33 |
+
|
| 34 |
+
## 我们的表现力评分工具:DeEAR
|
| 35 |
+
|
| 36 |
+
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](https://github.com/FreedomIntelligence/ExpressiveSpeech) 上找到它。
|
| 37 |
+
|
| 38 |
+
|
| 39 |
## 数据格式
|
| 40 |
|
| 41 |
数据集的组织结构如下:
|
|
|
|
| 54 |
```
|
| 55 |
|
| 56 |
- **`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:
|
| 57 |
+
- `audio-path`: 音频文件的相对路径。
|
| 58 |
- `value`: 由 ASR 生成的文本转录。
|
| 59 |
- `emotion`: 来自原数据集的情感标签。
|
| 60 |
- `expressiveness_scores`: 来自 **DeEAR** 模型的表现力得分。
|
| 61 |
|
| 62 |
+
## JSONL 文件示例
|
| 63 |
+
|
| 64 |
+
每行 JSONL 包含一个 `conversations` 字段,其中存放语音轮次数组。
|
| 65 |
+
示例:
|
| 66 |
+
|
| 67 |
+
```json
|
| 68 |
+
{"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"}]}
|
| 69 |
+
{"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"}]}
|
| 70 |
+
```
|
| 71 |
+
|
| 72 |
+
*注*: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。
|
| 73 |
+
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。
|
| 74 |
+
|
| 75 |
+
|
| 76 |
## 授权协议
|
| 77 |
|
| 78 |
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。
|