License:
CC-BY-SA-4.0
Steward:
CommunityDataset ID:
cmuwuxgta009d07mfzupkk4b0
Release Date: 10/6/2026
Format: FLAC, TSV
Size: 713.26 MB
This dataset contains 8,000 read Javanese utterances (12.7 hours, 786 speakers) with their original audio and transcripts, plus the full transcript file of the corpus (185,076 lines). It is a subset of OpenSLR 35, the "Large Javanese ASR training data set" collected by Google in collaboration with Reykjavik University and Universitas Gadjah Mada (CC BY-SA 4.0). The subset was used to train an Indonesian–Javanese ASR submission to the Lost in Transcription challenge (MDC / DrivenData, 2026), whose rules require external training data to be available on MDC. An added column gives the normalised transcripts used in training.
Licensing
Creative Commons Attribution Share Alike 4.0 International (CC-BY-SA-4.0)
https://spdx.org/licenses/CC-BY-SA-4.0.htmlAttribution and ShareAlike, as required by CC BY-SA 4.0. Please cite: Kjartansson, O., Sarin, S., Pipatsrisawat, K., Jansche, M., & Ha, L. (2018). Crowd-Sourced Speech Corpora for Javanese, Sundanese, Sinhala, Nepali, and Bangladeshi Bengali. Proc. SLTU 2018, 52–55.
None beyond the licence (CC BY-SA 4.0 does not permit additional restrictions).
Ethical Review
(i) Nature: read Javanese speech from the original OpenSLR 35 release, with real speakers' voices. Speakers are identified only by the anonymous IDs of the original release; there are no names or demographic data. (ii) Lawful basis: CC BY-SA 4.0 permits redistribution with attribution and share-alike. (iii) Permissions: the original publishers' CC BY-SA 4.0 release. Recording and consent were handled by the original data collectors (see Kjartansson et al., 2018). No new personal data was added. (iv) Safeguards: audio and transcripts are unchanged; no personal data was added; non-exclusive listing with attribution. We ask users not to attempt speaker identification (a request, not a licence condition).
Intended Use
Training and evaluating Javanese ASR; reproducing the Lost in Transcription competition submission.
This archive re-hosts, unchanged, the part of the OpenSLR 35 corpus that was used to train the Indonesian–Javanese
submission of participant inyongkhafid in the Lost in Transcription challenge (Mozilla Data Collective / DrivenData, 2026).
The competition requires external training data that is not already on MDC to be published on MDC; this upload fulfils that rule.
Original resource: Large Javanese ASR training data set (SLR35), https://www.openslr.org/35/ — "Javanese ASR training data set containing ~185K utterances", developed by Google in partnership with Reykjavik University and Universitas Gadjah Mada. License: Attribution-ShareAlike 4.0 International (CC BY-SA 4.0).
| Path | What |
|---|---|
audio/{xx}/{id}.flac | 8,000 utterances, original FLAC files from the OpenSLR release, not re-encoded (12.7 h, 786 speakers). A fixed random subset (seed 20260827) of the full corpus. |
utt_spk_text.tsv | Original transcript index of the full corpus (185,076 lines: FileID, UserID, transcription), unchanged. All of these transcripts (text only, no audio) were used for text-only adaptation of one model. |
subset_used.tsv | The audio utterances used: id, speaker, transcript (original), transcript_as_used (training normalisation, see below). |
LICENSE | CC BY-SA 4.0 legal code. |
SHA256SUMS | SHA-256 of every file in this archive. |
transcript_as_used): Unicode NFC; diacritics folded (è/é/ê/ě → e, ō → o, ā → a); typographic quotes → ASCII;
bracketed tags removed; dh → d, th → t (to match the competition's Javanese spelling); whitespace collapsed. The audio was not modified.Released under CC BY-SA 4.0, the licence of the original corpus. Changes made: selection of a subset, and one added column (transcript_as_used).
No other terms are added: CC BY-SA 4.0 does not permit additional restrictions.
Please cite the original corpus:
@inproceedings{kjartansson-etal-sltu2018,
title = {{Crowd-Sourced Speech Corpora for Javanese, Sundanese, Sinhala, Nepali, and Bangladeshi Bengali}},
author = {Oddur Kjartansson and Supheakmungkol Sarin and Knot Pipatsrisawat and Martin Jansche and Linne Ha},
booktitle = {Proc. The 6th Intl. Workshop on Spoken Language Technologies for Under-Resourced Languages (SLTU)},
year = {2018}, address = {Gurugram, India}, month = aug, pages = {52--55}
}
09fc6cb1a1 — wetone Nur kuwi Setu Pon0d867c3a26 — Victoria Beckham seneng ngrungokne musik rap12a9f95411 — Any insect or spider may often be called a bug593428d698 — Henry banjur dadi tentara lan oleh tugas pitung taunccc43ffe78 — Susu saka sawetara kéwan dikumpulaké manungsa kanggo diombéed7a93ff93 — Lance Armstrong kuwi alumni Stamford UniversityOriginal source: https://www.openslr.org/35/ (the corpus is also distributed there). Non-exclusive.
Legal basis: Google released OpenSLR 35 under the Creative Commons Attribution-ShareAlike 4.0 International licence (LICENSE file at https://www.openslr.org/35/). Section 2(a)(1) of this licence lets anyone reproduce and share the material, in whole or in part, including for commercial purposes, on two conditions: attribution (Section 3(a)) and ShareAlike (Section 3(b)). This upload meets both: it keeps the same licence, credits the original publisher (copyright notice, citation and link), and its audio and transcripts are unchanged (8,000 of the ~185,000 original utterances). Other platforms already redistribute this corpus under the same licence; for example, it is included in the "openslr" dataset on Hugging Face, which was not published by us. On 6 October 2026, at the request of the MDC reviewers, we notified the original authors (Google) that this copy is published on MDC under the same licence.
You might also like...
Dataset attributes