README.md

metadata

language:
  - kas
license: cc-by-4.0
task_categories:
  - text-to-speech
pretty_name: Kashmiri TTS Dataset
dataset_info:
  features:
    - name: audio
      dtype: audio
    - name: text
      dtype: string
  splits:
    - name: train
      num_bytes: 8934131482
      num_examples: 33182
    - name: test
      num_bytes: 563990590
      num_examples: 2272
    - name: validation
      num_bytes: 1135150113
      num_examples: 4542
  download_size: 12692057127
  dataset_size: 10633272185
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*
      - split: validation
        path: data/validation-*

Kashmiri TTS Dataset

Project Page | GitHub | Paper

Overview

This dataset is the Text-to-Speech (TTS) corpus for the Kashmiri language, as presented in the paper "Bolbosh: Script-Aware Flow Matching for Kashmiri Text-to-Speech".

The dataset is a derived and curated combination of Kashmiri speech data from the IndicVoices-R corpus and the RASA speech dataset. It was used to develop Bolbosh, an open-source neural TTS system designed to handle the specific orthographic and phonological challenges of the Kashmiri language.

---

Data Sources and Attribution

This dataset is derived from the following resources:

IndicVoices-R

• A large-scale multilingual speech corpus covering several Indic languages, including Kashmiri.
• Source: AI4Bharat/IndicVoices
• License: CC BY 4.0

RASA (Resource for Audio-visual Speech Analysis)

• A speech dataset supporting research in speech and language technologies.
• License: As provided by the original RASA dataset distribution.

Note: This dataset does not replace the original datasets. Users are encouraged to cite and respect the licenses of the original sources.

---

Dataset Structure

The dataset follows the Hugging Face audiofolder format, enabling automatic audio playback and preview in the dataset viewer.

Audio Specifications

• Sampling Rate: 22,050 Hz
• Format: mono (derived from studio and field recordings)

Splits

Split	Number of Utterances
train	33,182
validation	4,542
test	2,272

Citation

If you use this dataset or the Bolbosh system, please cite the following paper:

@article{ashraf2024bolbosh,
  title={Bolbosh: Script-Aware Flow Matching for Kashmiri Text-to-Speech},
  author={Ashraf, Tajamul and Zargar, Burhaan Rasheed and Muizz, Saeed Abdul and Mushtaq, Ifrah and Mehdi, Nazima and Gillani, Iqra Altaf and Kak, Aadil Amin and Bashir, Janibul},
  journal={arXiv preprint arXiv:2603.07513},
  year={2024}
}