KAI-indian-emotional-speech-corpus

# Indian Emotional Speech Corpus

## Dataset Description

This dataset comprises high-quality audio recordings of Indian speakers reading a standardized 50-word paragraph in four distinct emotional tones — **happy**, **sad**, **surprised**, and **angry**.

Each recording is approximately 20–25 seconds long and includes the full paragraph with tone shifts at specific points.

**Text spoken by all participants:**

> (happy tone) *Last Monday was perfect—I got the job I’d been dreaming of! I screamed, hugged my parents, and we even had cake for breakfast.*
> (sad tone) *But two days later, my mom called—our dog Bruno had passed away in his sleep. I didn’t even get to say goodbye.*
> (surprised tone) *And just when I thought the week couldn’t get any crazier, I got a surprise call—I was selected for a global fellowship I never even applied for!*
> (angry tone) *But earlier that day, my Wi-Fi cut out again during an important meeting... third time this week! It’s beyond frustrating. What a week. Honestly, I’m still processing it all.*

The dataset supports training and evaluation of models in:

- Automatic Speech Recognition (ASR)
- Emotional tone classification
- Voice synthesis and generation
- Emotion-aware conversational agents

---

## Dataset Structure

- **Audio Files**: Audio clips are stored in the `audio/` directory with names like `audio_001.wav`, `audio_002.mp3`, etc.
- **Formats**: `.wav`, `.mp3`, `.m4a`
- **Duration**: Each file is 20–25 seconds long

- **Metadata** (in `metadata.csv`):
- `file_name`: Name of the audio file
- `transcription`: Full 50-word paragraph
- `emotional_tones`: Order of tones (e.g., `happy;sad;surprised;angry`)
- `age_group`: e.g., `18–30`, `31–45`, `46–60`
- `gender`: `male` or `female`
- `region`: e.g., `North India`, `South India`, `East India`, `West India`

---

## Intended Uses

### ✅ Direct Use

- Training and benchmarking ASR models with Indian-accented English
- Emotion detection and classification from voice
- Research in affective computing and empathetic AI

### ❌ Out-of-Scope Use

- Real-time or production-grade systems
- Commercial use without proper CC BY 4.0 attribution
- Clinical or diagnostic use cases

---

## Considerations and Limitations

- ❗ The dataset is small (<1,000 samples) and not fully representative of India's linguistic and emotional diversity
- 💡 Emotions are subjective — classification results may vary by listener or model
- 🔄 Future versions will aim to expand multilingual support and speaker diversity

---

## License

**CC BY 4.0** — You can use, modify, and share the dataset with appropriate credit.

---

## Contact

- For queries or collaborations related to datasets, contact at :
- [e-mail]
- [e-mail]

---

## Citation
**BibTeX:**
```bibtex
@misc{indian_emotional_speech_corpus,
title = {Indian Emotional Speech Corpus},
author = {Contributors from across India},
year = {2025},
howpublished = {\url{https://huggingface.co/datasets/indian_emotional_speech_corpus}},
note = {Dataset available under CC-BY-4.0 License}
}

# 印度情感语音语料库（Indian Emotional Speech Corpus）

## 数据集描述

本数据集包含印度说话者朗读标准化50词段落的高质量音频录音，共涵盖四种明确的情感语调——**开心**、**悲伤**、**惊讶**与**愤怒**。

每条录音时长约20至25秒，完整朗读该段落，并在指定位置完成语调转换。

**所有参与者朗读的文本如下：**

> （开心语调）*上周一完美至极——我拿到了梦寐以求的工作！我尖叫着拥抱了父母，甚至还把蛋糕当早餐享用。*
> （悲伤语调）*但两天后，我妈妈打来电话——我的狗狗布鲁诺在睡梦中离世了。我甚至都没能和它道别。*
> （惊讶语调）*就在我以为这周不会再发生更离谱的事时，我接到了一个惊喜来电——我入选了一项我从未申请过的全球奖学金项目！*
> （愤怒语调）*但当天早些时候，我的Wi-Fi在一场重要会议期间又断了——这是本周第三次了！简直令人抓狂。这周真是糟透了。说实话，我还在消化这一切。*

本数据集可支持以下场景下的模型训练与评估：
- 自动语音识别（ASR）
- 情感语调分类
- 语音合成与生成
- 情感感知对话智能体

## 数据集结构

- **音频文件**：音频片段存储于`audio/`目录下，命名格式如`audio_001.wav`、`audio_002.mp3`等。
- **文件格式**：支持`.wav`、`.mp3`、`.m4a`三种格式
- **时长**：每条音频文件时长为20至25秒

- **元数据**（存储于`metadata.csv`中）：
- `file_name`：文件名
- `transcription`：完整转录文本
- `emotional_tones`：语调顺序（例如`happy;sad;surprised;angry`）
- `age_group`：年龄段（例如`18–30`、`31–45`、`46–60`）
- `gender`：性别，可选`male`（男性）或`female`（女性）
- `region`：所属地区（例如`North India`（印度北部）、`South India`（印度南部）、`East India`（印度东部）、`West India`（印度西部））

## 预期使用场景

### ✅ 合法使用范围
- 针对印度口音英语的自动语音识别模型训练与基准测试
- 语音情感检测与分类
- 情感计算与共情人工智能领域的研究

### ❌ 禁止使用范围
- 实时或生产级系统
- 未遵循CC BY 4.0协议进行适当署名的商业使用
- 临床或诊断类应用场景

## 注意事项与局限性

❗ 数据集规模较小（样本量不足1000），未能完全覆盖印度的语言与情感多样性
💡 情感表达具有主观性——分类结果可能因听众或模型不同而存在差异
🔄 未来版本将致力于拓展多语言支持与说话者多样性

## 授权协议

**CC BY 4.0** — 您可在恰当署名的前提下使用、修改与共享本数据集。

## 联系方式

- 若有数据集相关咨询或合作需求，请联系：
- [e-mail]
- [e-mail]

## 引用格式

**BibTeX:**
bibtex
@misc{indian_emotional_speech_corpus,
title = {Indian Emotional Speech Corpus},
author = {Contributors from across India},
year = {2025},
howpublished = {url{https://huggingface.co/datasets/indian_emotional_speech_corpus}},
note = {Dataset available under CC-BY-4.0 License}
}