severo/speech-rj-hi

---
dataset_info:
features:
- name: audio
dtype: audio
- name: sentence
dtype: string
splits:
- name: train
num_bytes: 3672926800.4989805
num_examples: 422603
- name: test
num_bytes: 36510981.394019544
num_examples: 4269
download_size: 2808288472
dataset_size: 3709437781.893
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
license: mit
task_categories:
- text-to-speech
- automatic-speech-recognition
language:
- hi
pretty_name: Rajasthani Speech Dataset
size_categories:
- 100K<n<1M
---
# Rajasthani Hindi Speech Dataset

<!-- Provide a quick summary of the dataset. -->
This dataset consists of audio recordings of participants reading out stories in Rajasthani Hindi, one sentence at a time. We had 98 participants from Soda, Rajasthan. Each participant read 30 stories. In total, we have 426873 recordings in this dataset. We had roughly 58 male participants and 40 female participants.

> **Point to Note:**
> While random sampling suggests that most users have to their best effort tried to accurately read out the sentences, we have not performed any quality analysis on the data. There could be errors in some of the recordings.

<!-- Provide a longer summary of what this dataset is. -->

### Dataset Sources

<!-- Provide the basic links for the dataset. -->

- **Link:** [Download](https://www.microsoft.com/en-gb/download/details.aspx?id=105385)
- **Curated By:** [Kalika Bali](https://www.microsoft.com/en-us/research/people/kalikab/downloads/)

## Dataset Structure

<!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->
Contains two headers: audio and sentence containing the Audio file and sentence respectively.

原始信息汇总

Rajasthani Hindi Speech Dataset

数据集概述

该数据集包含参与者在拉贾斯坦语（Rajasthani Hindi）中逐句朗读故事的音频记录。共有98名来自Soda, Rajasthan的参与者，每位参与者朗读了30个故事。总计有426873条录音，其中约58名男性参与者和40名女性参与者。

数据集结构

数据集包含两个字段：

• audio: 音频文件
• sentence: 句子文本

数据分割

数据集分为训练集和测试集：

• train: 包含422603个样本，总大小为3672926800.4989805字节
• test: 包含4269个样本，总大小为36510981.394019544字节

数据集大小

• 下载大小：2808288472字节
• 数据集总大小：3709437781.893字节

配置

• 默认配置（default）包含训练集和测试集的数据文件路径：
  • 训练集路径：data/train-*
  • 测试集路径：data/test-*

许可证

MIT许可证

任务类别

• 文本到语音转换
• 自动语音识别

语言

• 拉贾斯坦语（hi）

数据集名称

Rajasthani Speech Dataset

数据集规模

100K<n<1M