README.md

metadata

dataset_info:
  - config_name: audiobooks
    features:
      - name: audio
        dtype: audio
      - name: text
        dtype: string
      - name: duration
        dtype: float64
      - name: id
        dtype: string
      - name: title
        dtype: string
      - name: lang
        dtype: string
    splits:
      - name: train
  - config_name: audiobooks_rus_translations
    features:
      - name: audio
        dtype: audio
      - name: text
        dtype: string
      - name: duration
        dtype: float64
      - name: id
        dtype: string
      - name: rus_title
        dtype: string
      - name: title
        dtype: string
      - name: lang
        dtype: string
    splits:
      - name: train
configs:
  - config_name: audiobooks
    data_files:
      - split: train
        path: train*.parquet
  - config_name: audiobooks_rus_translations
    data_files:
      - split: train
        path: rus_train*.parquet
language:
  - tt
  - cv
  - udm
  - mdf
  - myv
  - mhr
  - ru
pretty_name: 'Volga Fairytales TTS '
task_categories:
  - text-to-speech
  - automatic-speech-recognition
  - audio-to-audio
tags:
  - audio
  - tts
  - speech
  - speech-to-speech
  - translation
size_categories:
  - 1K<n<10K

Volga Fairytales TTS Dataset

Dataset Summary

Volga Fairytales TTS is a speech dataset containing audio recordings of traditional fairytales in several languages of the Volga region.

The original recordings were sourced from this site.

The dataset comprises recordings of fairytale texts read by professional actors. The dataset is intended for text-to-speech (TTS) research and development in Turkic and other regional languages including Tatar, Chuvash, Mari, Udmurt, Erzya and Moksha. The dataset also includes Russian translations/recordings, but they are not aligned on a sentence level.

Dataset Structure

Parts:
The dataset contains two main configurations:

• audiobooks: Contains the original fairytale recordings in their native languages.
• audiobooks_rus_translations: Contains the recordings of Russian translations for original fairytales.

Data Fields:

• audio: The audio file containing the spoken fairytale segment.
• text: The corresponding transcription.
• duration: The duration of the audio segment in seconds.
• id: A unique identifier for each segment.
• title: The title of the fairytale.
• lang: The language code of the recording (e.g., "tat" for Tatar).
• rus_title: (Only in audiobooks_rus_translations) The Russian title of the fairytale.

Data Processing

Text Processing:

• All the text was splited, using razdel library.

Audio Processing:

• All audio was aligned to the text as each word had timestamps.

Technical Details:

Dataset Type	speech corpus for TTS
Language	Tatar, Chuvash, Meadow–Eastern Mari, Udmurt, Erzya, Moksha, Russian
Speech Style	scripted monologue
Content	fairytales
Audio Parameters	22.05 kHz, 32 bits, mono
File Format	WAV (PCM) TXT (UTF-8)
Recording Environment	quiet indoor environment + background music

Durations:

Language	Duration	Utterances
Tatar	0.74 hours	355
Chuvash	0.82 hours	425
Meadow–Eastern Mari	0.57 hours	384
Udmurt	0.23 hours	147
Erzya	0.20 hours	146
Moksha	0.36 hours	164
Russian	2.46 hours	1464

Usage Considerations

• The dataset includes a variety of fairytales that showcase cultural and linguistic aspects of the Volga region.
• The dataset is suitable for training text-to-speech systems, automatic speech recognition, and speech-to-speech translation systems.
• Researchers should be aware of potential variations in speaking styles and background audio conditions.