This dataset consists of raw, unlabeled audio recordings primarily in the Chuvash language. It is aggregated from various sources including audiobooks, religious texts, and lyrical content. It is intended for self-supervised learning, pre-training automatic speech recognition (ASR) models, or audio analysis tasks.

Note: While the primary language is Chuvash, some recordings may contain segments of Russian speech or code-switching, reflecting the bilingual nature of the region.

Dataset Structure & Statistics

The dataset is divided into several subsets. Below is the summary of the data distribution:

Subset Name	Description	Files	Hours	Seconds
lirics	Songs / Lyrical content	4,195	3.17	11,421.79
books_part_1	Audiobooks (Part 1)	9,481	7.28	26,193.70
books_part_2	Audiobooks (Part 2)	9,043	8.63	31,067.10
bible	Religious texts	24,437	13.88	49,980.20
bib	General Audiobooks Collection	323,820	186.08	669,870.52
TOTAL	All Subsets Combined	370,976	219.04	788,533.31

Audio Specifications

Format: Flac

Length Distributions by Subset

Below are the duration distributions for each subset of the data.

1. Subset: Lirics

2. Subset: Books (Parts 1 & 2)

3. Subset: Bible

4. Subset: Bib (Large Collection of audiobooks)

(Diverse audiobooks)

Usage

This dataset does not contain transcriptions. It is suitable for:

Self-supervised pre-training (e.g., Wav2Vec 2.0, HuBERT, XLSR)
Language identification (Chuvash vs. Russian)
Voice Activity Detection (VAD) analysis
Prosody and speech synthesis research

License

Creative Commons Attribution Non Commercial 4.0

License and acceptable use

License: CC BY-NC 4.0 — free for research and other non-commercial use.
Commercial use: possible under a separate agreement with the SpeechCollector community — reach out on Telegram: @Michaelya.
Surveillance: please do not use this dataset to build systems for surveillance, covert monitoring, or tracking of individuals.

Project website (ru / en / cv): https://chuvash-asr.vercel.app

Downloads last month: 19

Total file size:

38.2 GB