Loading datasets...
Task: ASR
Release Date: 3/20/2026
Format: MP3
Size: 194.33 MB
Share
A collection of spontaneous responses to questions in Kelabit (kzi).
Restrictions/Special Constraints
None provided.
Forbidden Usage
It is forbidden to attempt to determine the identity of speakers in the Common Voice datasets. It is forbidden to re-host or re-share this dataset.
Intended Use
This dataset is intended to be used for training and evaluating automatic speech recognition (ASR) models. It may also be used for applications relating to computer-aided language learning (CALL) and language or heritage revitalisation.
kzi)This datasheet is for sps-corpus-3.0-2026-03-09 of the Mozilla Common Voice Spontaneous Speech dataset for Kelabit [kzi - kzi]. The dataset contains 1717 clips representing 9.51 hours of recorded speech (0.22 hours validated) from 21 speakers.
Kelabit is an Austronesian language in the Apo Duat subgroup. It is spoken by the Kelabit people and is indigenous to Sarawak, Malaysia and North Kalimantan, Indonesia.
The dataset clips are categorised by transcription status and training-set assignment. The following tables summarise the distribution.
| Bucket | Clips | % |
|---|---|---|
| Transcribed & Validated | 59 | 3.4% |
| Transcribed & Pending | 1,658 | 96.6% |
| Not transcribed | 0 | 0.0% |
| Bucket | Clips | % |
|---|---|---|
| Train | 0 | 0.0% |
| Dev | 0 | 0.0% |
| Test | 0 | 0.0% |
| Unassigned | 1,717 | 100.0% |
Training split coverage: 0 of 59 transcribed & validated clips (0.0%)
| Bucket | Clips | % |
|---|---|---|
| Validated | 59 | 3.4% |
| Pending | 1,658 | 96.6% |
| Edited | 126 | 7.3% |
The Kelabit writing system is based on the Latin alphabet.
a e i o u
aa ee ii oo uu
b bp c d dt g gk h j k l m n ng p r s t w y ʼ
There follows a randomly selected sample of questions used in the corpus.
Nun bang terawe iko lem ayu guna bank lem talian?
Lem linuh iko. nun perkhidmatan nuk leng doo/perlu ngan lemulun mula?
Nun bada iko kinuan ja nuk mudeng mado ngan lun ruyung iyeh?
Kapeh ngutang nuk siit kerab nuuh idah desur ngepuun naru dagang?
Nun nuk midih iko rapang/tuseh ngerawe lem ayu talipaun?
There follows a randomly selected sample of transcribed responses from the corpus.
Luk tu'en uih mada' sah kerja kerja luk senaru' uih baru-baru inih iyeh neh nulung [err] nekap barang tu'en merey ngen lun merar.
Ah inan kanuh nuk tuen Lun Tauh iyeh ineh kanut [uhm] lun delaih, kanut lun desur dah kanut kadang. Dan juga' inan teh [uhm] music Sapeh ruyung kanut neh.
Nun program beken, program raut sape' unih sape' teh doo' pedingeren narih lem kampong nangey doo' teh [err] ngebpo erur narih lem muli' let lem lati'.
[cough]... Lem ayu' raut nih uih doo' pian raut ngen... ngen lem group lah. [sepeh...] kayu' [aah] raut betminten dih. Pruyung teh narih raut, pruyung teh narih [ngg...] naru' memukul dih lah kadi' narih doo' pian. Doo' teh pian narih raut pruyung ngen kawan narih. Inan the semanget muh ... nekaruh ngedeh men. Ineh neh doo' dih.
Nukanan nuk lang do,lam burul luk an narih kuman iylah nubah,lawih, labo nuk ruh burul narih kail.
The questions use the ’ character, while the responses use the ' character instead. These characters should be treated as equivalent.
Each row of a tsv file represents a single audio clip, and contains the following information:
client_id - hashed UUID of a given user
audio_id - numeric id for audio file
audio_file - audio file name
duration_ms - duration of audio in milliseconds
prompt_id - numeric id for prompt
prompt - question for user
transcription - transcription of the audio response
votes - number of people that who approved a given transcript
age - age of the speaker1
gender - gender of the speaker1
language - language name
split - for data modelling, which subset of the data does this clip pertain to
char_per_sec - how many characters of transcription per second of audio
quality_tags - some automated assessment of the transcription--audio pair, separated by |
transcription-length - character per second under 3 characters per second
speech-rate - characters per second over 30 characters per second
short-audio - audio length under 2 seconds
long-audio - audio length over 5 minutes
This dataset was partially funded by the Open Multilingual Speech Fund managed by Mozilla Common Voice.
This dataset is released under the Creative Commons Zero (CC-0) licence. By downloading this data you agree to not determine the identity of speakers in the dataset.
For a full list of age, gender, and accent options, see the demographics spec. These will only be reported if the speaker opted in to provide that information. ↩ ↩2