MERaLiON/sea_audiobench_datasets_ASR

SEA-SpeechBench — ASR (Automatic Speech Recognition)

This dataset is the automatic-speech-recognition (ASR) task of SEA-SpeechBench, a large-scale multitask benchmark for speech understanding across Southeast Asia. It contains 26,863 evaluation examples across eleven languages, drawn from fifteen source corpora, each pairing an audio recording with an instruction and a reference transcript.

Given the recording and the instruction, a model must transcribe the speech.

Quick start

Requires datasets>=4.0, which decodes audio through torchcodec and needs a system FFmpeg (versions 4–7):

pip install "datasets[audio]"

from datasets import load_dataset

ds = load_dataset(
    "MERaLiON/sea_audiobench_datasets_ASR",
    "asr_fleurs_th_30",
    split="train",
)

row = ds[0]
print(row["instruction"])
print(row["answer"])
print(row["language"], row["audio_length"])

# `context` is a torchcodec AudioDecoder, not a dict
samples = row["context"].get_all_samples()
waveform = samples.data          # torch.Tensor, shape (num_channels, num_samples)
sr = samples.sample_rate         # 16000
print(waveform.shape, sr)

If you need a numpy array — most feature extractors take one — squeeze the channel dimension (all ASR audio is mono):

audio = samples.data.squeeze(0).numpy()

Subsets

All data lives in a single train split and is intended for evaluation only. Config names follow asr_{source}[_{language}]_30.

Config	Source	Lang	Examples	Audio (hr)
asr_bloomspeech_en_30	Bloom-Speech	en	449	0.7
asr_bloomspeech_my_30	Bloom-Speech	my	50	0.1
asr_bloomspeech_tl_30	Bloom-Speech	tl	50	0.1
asr_bud500_30	Bud500	vi	1,000	0.7
asr_cv21_en_30	Common Voice 21	en	1,000	1.7
asr_cv21_id_30	Common Voice 21	id	1,000	1.1
asr_cv21_ta_30	Common Voice 21	ta	1,000	1.6
asr_cv21_th_30	Common Voice 21	th	1,000	1.4
asr_cv21_vi_30	Common Voice 21	vi	1,000	1.0
asr_cv21_zh_30	Common Voice 21	zh	1,000	1.7
asr_esd_en_30	ESD	en	1,000	0.8
asr_esd_zh_30	ESD	zh	1,000	0.9
asr_fleurs_id_30	FLEURS	id	684	2.3
asr_fleurs_km_30	FLEURS	km	765	3.1
asr_fleurs_lo_30	FLEURS	lo	399	1.3
asr_fleurs_ms_30	FLEURS	ms	749	2.3
asr_fleurs_my_30	FLEURS	my	870	3.7
asr_fleurs_th_30	FLEURS	th	1,000	3.3
asr_fleurs_tl_30	FLEURS	tl	928	4.4
asr_fleurs_vi_30	FLEURS	vi	850	2.9
asr_fleurs_zh_30	FLEURS	zh	944	3.1
asr_malcsc_30	ASR-MalCSC	ms	1,000	0.7
asr_mig_my_30	MIG	my	140	0.1
asr_openslr_km_30	OpenSLR	km	1,000	1.4
asr_openslr_my_30	OpenSLR	my	1,000	1.6
asr_openslr_ta_30	OpenSLR	ta	1,000	1.7
asr_sfdusc_30	ASR-SFDuSC	tl	1,000	1.2
asr_sg_streets_utterance_30	SG Streets	en	492	0.7
asr_sgpccsc_utterance_30	ASR-SgpCCSC	zh	1,000	0.9
asr_smaldusc_30	ASR-SMalDuSC	ms	1,000	2.1
asr_thai_elderly_th_30	Thai Elderly Speech	th	992	1.4
asr_thai_lotus_30	LOTUS	th	501	1.2
asr_vietmed_vi_30	VietMed	vi	1,000	1.7
Total			26,863	~52.9

Data fields

Field	Type	Description
context	Audio(sampling_rate=16000)	The audio recording.
instruction	string	The instruction given to the model.
answer	string	The reference transcript.
audio_length	float64	Duration of context, in seconds.
language	string	Language code of the subset.

See the paper for the instruction and answer formats, how utterances were selected, and the evaluation protocol.

Source data

ASR introduces no new recordings. All audio comes from existing corpora; the contribution is the utterance selection, instructions, and reference answers.

Source	Lang	Upstream license
Bloom-Speech	en, my, tl	CC BY-NC 4.0
Bud500	vi	Apache-2.0
Common Voice 21	en, id, ta, th, vi, zh	MPL 2.0
ESD	en, zh	MIT
FLEURS	id, km, lo, ms, my, th, tl, vi, zh	CC BY 4.0
ASR-MalCSC	ms	CC BY-NC-ND 4.0
MIG	my	Not specified
OpenSLR	km, my, ta	CC BY-SA 4.0
ASR-SFDuSC	tl	CC BY-NC-ND 4.0
SG Streets	en	Not specified
ASR-SgpCCSC	zh	CC BY-NC-ND 4.0
ASR-SMalDuSC	ms	CC BY-NC-ND 4.0
Thai Elderly Speech	th	CC BY-SA 4.0
LOTUS	th	CC BY-NC-SA 4.0
VietMed	vi	MIT

Citation

If you use this dataset, please cite our benchmark and the source corpora.

This benchmark

@inproceedings{liao2026seaspeechbench,
  title     = {{SEA-SpeechBench}: A Large-Scale Multitask Benchmark for
               Speech Understanding Across Southeast Asia},
  author    = {Liao, Jingyi and Zhang, Wenyu and Liu, Zhuohan and
               He, Yingxu and Lin, Geyu and Zou, Xunlong and Sun, Shuo and
               Alsagoff, Syed Ali Redha and Aw, Ai Ti},
  booktitle = {Proceedings of the 2026 Conference on Empirical Methods in
               Natural Language Processing (EMNLP)},
  year      = {2026}
}

Source corpora

@inproceedings{bloomspeech,
  title     = {Bloom Library: Multimodal Datasets in 300+ Languages for a
               Variety of Downstream Tasks},
  author    = {Leong, Colin and Nemecek, Joshua and Mansdorfer, Jacob and
               Filighera, Anna and Owodunni, Abraham and Whitenack, Daniel},
  booktitle = {Conference on Empirical Methods in Natural Language Processing},
  year      = {2022}
}

@misc{Bud500,
  title  = {Bud500: A comprehensive Vietnamese ASR dataset},
  author = {Pham, Anh and Tran, Khanh Linh and Nguyen, Linh and
            Cao, Thanh Duy and Phan, Phuc and Nguyen, Duong A},
  year   = {2024},
  url    = {https://github.com/quocanh34/Bud500}
}

@inproceedings{commonvoice,
  title     = {Common Voice: A Massively-Multilingual Speech Corpus},
  author    = {Ardila, Rosana and Branson, Megan and Davis, Kelly and
               Kohler, Michael and Meyer, Josh and Henretty, Michael and
               Morais, Reuben and Saunders, Lindsay and Tyers, Francis and
               Weber, Gregor},
  booktitle = {Conference on Language Resources and Evaluation},
  year      = {2020}
}

@article{esd,
  title     = {Emotional voice conversion: Theory, databases and ESD},
  author    = {Zhou, Kun and Sisman, Berrak and Liu, Rui and Li, Haizhou},
  journal   = {Speech Communication},
  volume    = {137},
  pages     = {1--18},
  year      = {2022},
  publisher = {Elsevier}
}

@inproceedings{fleurs,
  title     = {FLEURS: Few-shot Learning Evaluation of Universal
               Representations of Speech},
  author    = {Conneau, Alexis and Ma, Min and Khanuja, Simran and
               Zhang, Yu and Axelrod, Vera and Dalmia, Siddharth and
               Riesa, Jason and Rivera, Clara and Bapna, Ankur},
  booktitle = {IEEE Spoken Language Technology Workshop},
  year      = {2022}
}

@misc{malcsc,
  title        = {ASR-MalCSC: Malay Conversational Speech Corpus},
  author       = {{Magic Data Technology}},
  year         = {2025},
  howpublished = {Online: MagicHub},
  url          = {https://magichub.com/datasets/malay-conversational-speech-corpus/}
}

@misc{mig,
  title        = {mig-burmese-audio-transcription},
  author       = {{Myanmar Innovative Group}},
  year         = {2025},
  howpublished = {\url{https://huggingface.co/datasets/Ko-Yin-Maung/mig-burmese-audio-transcription}}
}

@inproceedings{OpenSLR1,
  title     = {{A Step-by-Step Process for Building TTS Voices Using Open
               Source Data and Framework for Bangla, Javanese, Khmer, Nepali,
               Sinhala, and Sundanese}},
  author    = {Sodimana, Keshan and Pipatsrisawat, Knot and Ha, Linne and
               Jansche, Martin and Kjartansson, Oddur and De Silva, Pasindu
               and Sarin, Supheakmungkol},
  booktitle = {Proc. The 6th Intl. Workshop on Spoken Language Technologies
               for Under-Resourced Languages (SLTU)},
  pages     = {66--70},
  month     = aug,
  year      = {2018},
  address   = {Gurugram, India},
  url       = {https://dx.doi.org/10.21437/SLTU.2018-14}
}

@inproceedings{OpenSLR2,
  title     = {{Open-source Multi-speaker Speech Corpora for Building
               Gujarati, Kannada, Malayalam, Marathi, Tamil and Telugu
               Speech Synthesis Systems}},
  author    = {He, Fei and Chu, Shan-Hui Cathy and Kjartansson, Oddur and
               Rivera, Clara and Katanova, Anna and Gutkin, Alexander and
               Demirsahin, Isin and Johny, Cibu and Jansche, Martin and
               Sarin, Supheakmungkol and Pipatsrisawat, Knot},
  booktitle = {Proceedings of The 12th Language Resources and Evaluation
               Conference (LREC)},
  pages     = {6494--6503},
  month     = may,
  year      = {2020},
  address   = {Marseille, France},
  publisher = {European Language Resources Association (ELRA)},
  url       = {https://www.aclweb.org/anthology/2020.lrec-1.800}
}

@inproceedings{OpenSLR3,
  title     = {{Burmese Speech Corpus, Finite-State Text Normalization and
               Pronunciation Grammars with an Application to Text-to-Speech}},
  author    = {Oo, Yin May and Wattanavekin, Theeraphol and Li, Chenfang and
               De Silva, Pasindu and Sarin, Supheakmungkol and
               Pipatsrisawat, Knot and Jansche, Martin and
               Kjartansson, Oddur and Gutkin, Alexander},
  booktitle = {Proceedings of The 12th Language Resources and Evaluation
               Conference (LREC)},
  month     = may,
  year      = {2020}
}

@misc{sfdusc_ph,
  title        = {ASR-SFDuSC: A Scripted Filipino Daily-use Speech Corpus},
  author       = {{Magic Data Technology}},
  year         = {2023},
  howpublished = {MagicHub},
  url          = {https://magichub.com/datasets/filipino-scripted-speech-corpus-daily-use-sentence/}
}

@inproceedings{sg_streets,
  title     = {Enriching Rare Word Representations in Neural Language Models
               by Embedding Matrix Augmentation},
  author    = {Khassanov, Yerbolat and Zeng, Zhiping and Pham, Van Tung and
               Xu, Haihua and Chng, Eng Siong},
  booktitle = {Interspeech 2019},
  pages     = {3505--3509},
  month     = sep,
  year      = {2019},
  publisher = {ISCA},
  doi       = {10.21437/Interspeech.2019-1858}
}

@misc{sgpccsc,
  title        = {{ASR-SgpCCSC}},
  author       = {{Magic Data Technology}},
  howpublished = {\url{https://magichub.com/datasets/singaporean-chinese-conversational-speech-corpus/}}
}

@misc{SMalDuSC,
  title        = {{ASR-SMalDuSC}: A Scripted Malay Daily-use Speech Corpus},
  author       = {{Magic Data Technology}},
  howpublished = {\url{https://magichub.com/datasets/malay-scripted-speech-corpus-daily-use-sentence/}},
  year         = {2023}
}

@misc{thaielderly,
  title        = {Thai Elderly Speech dataset by Data Wow and VISAI},
  author       = {{VISAI AI Company Limited} and {Data Wow Company Limited}},
  year         = {2022},
  howpublished = {\url{https://github.com/VISAI-DATAWOW/Thai-Elderly-Speech-dataset/releases/tag/v1.0.0}}
}

@inproceedings{lotus,
  title     = {LOTUS-BN: A Thai broadcast news corpus and its research
               applications},
  author    = {Chotimongkol, Ananlada and Saykhum, Kwanchiva and
               Chootrakool, Patcharika and Thatphithakkul, Nattanun and
               Wutiwiwatchai, Chai},
  booktitle = {2009 Oriental COCOSDA International Conference on Speech
               Database and Assessments},
  pages     = {44--50},
  year      = {2009},
  doi       = {10.1109/ICSDA.2009.5278377}
}

@inproceedings{VietMed,
  title     = {VietMed: A Dataset and Benchmark for Automatic Speech
               Recognition of Vietnamese in the Medical Domain},
  author    = {Le-Duc, Khai},
  booktitle = {Proceedings of the Joint International Conference on
               Computational Linguistics, Language Resources and Evaluation
               (LREC-COLING 2024)},
  year      = {2024}
}