metadata

annotations_creators:
  - expert-generated
language_creators:
  - expert-generated
languages:
  - kk
license:
  - cc-by-nc-4.0
multilinguality:
  - monolingual
pretty_name: Kazakh Traditional Audio Archive
size_categories:
  - 1K<n<10K
source_datasets:
  - original
task_categories:
  - audio-classification
  - automatic-speech-recognition
  - text-to-speech
task_ids:
  - multi-class-audio-classification
  - automatic-speech-recognition
  - text-to-speech

Dataset Card for Kazakh Traditional Audio Archive

Dataset Summary

The Kazakh Traditional Audio Archive is a curated open dataset containing rare and historical audio recordings from Kazakh culture. The dataset includes traditional musical forms such as kui, zhyr/terme, folk and composed songs, lullabies, retro music, and audio fairy tales.

These audio files are designed for use in machine learning tasks including ASR (Automatic Speech Recognition), TTS (Text-to-Speech), music classification, and ethnomusicological research.

Supported Tasks and Leaderboards

Supported Tasks:

audio-classification: Classifying audio by genre or type (e.g., kui, zhyr, retro)
automatic-speech-recognition: Recognizing lyrics or speech in Kazakh
text-to-speech: Synthesizing culturally appropriate TTS models
audio-generation and audio-captioning (experimental)

Languages

The primary language of the dataset is Kazakh (kk).

Dataset Structure

Data Fields

Each entry in the dataset has the following metadata:

file: Relative path to audio file
title: Name of the piece
type: One of kui, zhyr_terme, an_kompozitor, ertegi, besik, retro
language: Always "kk" (Kazakh)
duration: Duration of the audio in seconds

Example:

{
  "file": "data/kui/kui_001.wav",
  "title": "Saryzhailau",
  "type": "kui",
  "language": "kk",
  "duration": 125.4
}

Data Splits

The dataset has a single train split (default for cultural datasets).

Dataset Creation

Source Data

The recordings were collected from public domain sources and cultural preservation projects. All materials are either:

Traditional public domain content
Released by contributors for non-commercial research and academic purposes

Annotations

Metadata annotations were manually created by linguistic and cultural experts.

Licensing

This dataset is licensed under the Creative Commons Attribution-NonCommercial 4.0 (CC BY-NC 4.0).
It may be used freely for research and non-commercial applications. Commercial use is prohibited.

Citation

If you use this dataset in your research or application, please cite as:

@misc{kazakh_audio_2025,
  title={Kazakh Traditional Audio Archive},
  author={Republican Television and Radio Corporation "Kazakhstan"},
  year={2025},
  howpublished={\url{https://huggingface.co/datasets/rtrk/kazakh-traditional-audio}},
  note={Accessed April 2025}
}

Contributions

Curated and published by Republican Television and Radio Corporation "Kazakhstan".
Pull requests and metadata improvements are welcome.