---
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.