Hugging Face's logo

Datasets:
MERaLiON
/
sea_audiobench_datasets_SpR

Tasks:
Audio Classification
Modalities:
Audio
Text
Formats:
parquet
Languages:
English
Chinese
Thai
Size:
1K - 10K
ArXiv:
Tags:
speech
audio-llm
speaker-recognition
benchmark
southeast-asia
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
2
Dataset Viewer
Auto-converted to Parquet Duplicate
instruction
stringclasses
1 value
answer
stringclasses
2 values
audio_length
float64
2.74
9.79
audio_length_1
float64
1.06
5.63
audio_length_2
float64
1.06
6.4
language_1
stringclasses
1 value
language_2
stringclasses
1 value
context_1
audioduration (s)
1.06
5.63
context_2
audioduration (s)
1.06
6.4
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
8.341375
4.16
4.181375
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
4.586688
1.792
2.794688
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
6.07575
2.737875
3.337875
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
6.173875
1.8645
4.309375
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
5.393813
3.153813
2.24
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
5.440375
3.3425
2.097875
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
6.630313
3.323125
3.307188
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
6.741375
3.285375
3.456
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
4.932188
3.0335
1.898688
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.32
2.08
2.24
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
6.611563
4.416
2.195563
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.793938
2.37575
2.418188
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.821375
2.709375
2.112
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
6.98825
3.297938
3.690313
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.245375
2.176
2.069375
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
6.537875
3.337875
3.2
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
3.776
1.536
2.24
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
4.826688
2.624
2.202688
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
8.82
3.98
4.84
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
4.864563
2.517875
2.346688
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.8
2.88
1.92
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
4.484563
2.986688
1.497875
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
6.221938
3.85
2.371938
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
5.205375
2.858688
2.346688
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
3.945625
1.977875
1.96775
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
5.169875
2.752
2.417875
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
4.5525
2.688
1.8645
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
5.973375
3.285375
2.688
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.435
1.60575
2.82925
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
4.570937
2.577438
1.9935
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.25575
1.977875
2.277875
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.074688
2.048
2.026688
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
5.141375
3.050688
2.090688
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
3.925375
1.685375
2.24
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.565938
2.275
2.290938
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
6.356
3.796
2.56
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
6.89075
3.093375
3.797375
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
5.6
2.24
3.36
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
7.08
3.23
3.85
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
6.6155
3.669375
2.946125
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
4.97925
2.339188
2.640063
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
8.064
3.904
4.16
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
4.697875
2.88
1.817875
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.336688
2.368938
1.96775
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.480063
2.090688
2.389375
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
5.845375
2.88
2.965375
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
5.427375
2.617063
2.810313
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
5.824063
2.965375
2.858688
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
6.826688
3.584
3.242688
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.96
2.24
2.72
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
5.696
3.136
2.56
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
4.935625
2.417875
2.51775
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
5.376
2.24
3.136
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
7.445375
4.16
3.285375
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
6.08
3.2
2.88
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.71575
2.797875
1.917875
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
5.21575
2.357875
2.857875
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
5.24025
2.642188
2.598063
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
5.672938
3.285375
2.387563
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.423375
2.159188
2.264188
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
5.61925
3.343688
2.275563
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
5.88675
2.537375
3.349375
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
6.074125
3.648
2.426125
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
5.712188
2.832875
2.879313
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.288
2.24
2.048
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.535375
2.174563
2.360813
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
5.120063
2.325375
2.794688
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.520875
1.998875
2.522
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.133563
2.160125
1.973438
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
8.661375
3.690688
4.970688
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
5.216375
2.4
2.816375
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
3.73575
2.017875
1.717875
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
5.66
2.87
2.79
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.903125
2.237438
2.665688
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
4.970688
2.410688
2.56
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
4.053375
2.24
1.813375
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
5.120062
2.389375
2.730688
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
3.99575
2.377875
1.617875
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
8.102688
3.9
4.202688
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
6.91925
2.737875
4.181375
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
5.964562
3.103875
2.860688
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
4.157563
2.159938
1.997625
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
9.408
5.632
3.776
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
6.869375
3.2
3.669375
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
6.705438
3.904
2.801438
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
6.316188
3.127313
3.188875
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
5.56225
2.276875
3.285375
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.416063
2.133375
2.282688
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.842813
2.75
2.092813
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
7.317375
3.114688
4.202688
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.876
2.541938
2.334063
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
5.145438
2.54275
2.602688
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
4.600063
2.642188
1.957875
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
5.504063
2.986688
2.517375
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
5.620563
2.484563
3.136
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
4.909688
2.666
2.243688
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
7.227875
4.49
2.737875
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
5.328375
2.773375
2.555
ta
ta
Are the speakers in the two audio clips the same?
Yes, the speakers are the same.
5.504
3.072
2.432
ta
ta
Are the speakers in the two audio clips the same?
No, the speakers are different.
5.141375
3.349375
1.792
ta
ta
End of preview. Expand in Data Studio

SEA-SpeechBench — SpkR (Speaker Recognition)

This dataset is the speaker-recognition (SpkR) task of SEA-SpeechBench, a large-scale multitask benchmark for speech understanding across Southeast Asia. It contains 8,827 evaluation examples across eight languages, each pairing two audio recordings with an instruction and a reference answer.

Given the two recordings and the instruction, a model must compare the speakers in them.

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_SpkR",
    "spr_esd_en_30",
    split="train",
)

row = ds[0]
print(row["instruction"])
print(row["answer"])
print(row["language_1"], row["language_2"])

# `context_1` and `context_2` are torchcodec AudioDecoders, not dicts
samples = row["context_1"].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 SpkR 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 spr_{source}[_{language}]_30.

Config Source Lang Examples Audio (hr)
spr_emota_ta_30 EmoTa ta 936 1.5
spr_esd_en_30 ESD en 1,000 1.5
spr_esd_zh_30 ESD zh 1,000 1.8
spr_mig_my_30 MIG my 1,000 1.4
spr_sfdusc_30 SFDUSC tl 1,000 2.4
spr_smaldusc_30 ASR-SMalDuSC ms 1,000 4.3
spr_thai_elderly_th_30 Thai Elderly Speech th 960 2.8
spr_thai_ser_30 THAI SER th 931 3.0
spr_voxvietnam_vi_30 VoxVietnam-O vi 1,000 2.4
Total 8,827 ~21.1

Data fields

Field Type Description
context_1 Audio(sampling_rate=16000) The first audio recording.
context_2 Audio(sampling_rate=16000) The second audio recording.
instruction string The instruction given to the model.
answer string The reference answer.
audio_length_1 float64 Duration of context_1, in seconds.
audio_length_2 float64 Duration of context_2, in seconds.
audio_length float64 Duration in seconds.
language_1 string Language code of context_1.
language_2 string Language code of context_2.

See the paper for the instruction and answer formats, how the recording pairs were constructed, and the evaluation protocol.

Source data

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

Source Lang Upstream license
EmoTa ta EACL
ESD en, zh MIT
MIG my Not specified
SFDUSC tl CC BY-NC-ND 4.0
ASR-SMalDuSC ms CC BY-NC-ND 4.0
Thai Elderly Speech th CC BY-SA 4.0
THAI SER th CC BY-SA 4.0
VoxVietnam-O vi CC

Citation

If you use SpkR, 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 

@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{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{emota,
  title     = {EmoTa: A Tamil Emotional Speech Dataset},
  author    = {Thevakumar, Jubeerathan and Thavarasa, Luxshan and
               Sivatheepan, Thanikan and Kugarajah, Sajeev and
               Thayasivam, Uthayasanker},
  booktitle = {Proceedings of the First Workshop on Challenges in Processing
               South Asian Languages (CHiPSAL 2025)},
  pages     = {193--201},
  month     = {January},
  year      = {2025},
  address   = {Abu Dhabi, UAE},
  publisher = {International Committee on Computational Linguistics},
  url       = {https://aclanthology.org/2025.chipsal-1.19/}
}

@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}
}

@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}}
}

@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}}
}

@article{thaiser,
  title   = {THAI Speech Emotion Recognition (THAI-SER) corpus},
  author  = {Wongpithayadisai, Jilamika and Chaksangchaichot, Chompakorn and
             Sangnark, Soravitt and Prakrankamanant, Patawee and
             Gangwanpongpun, Krit and Boonpunmongkol, Siwa and
             Milindasuta, Premmarin and Na-Pombejra, Dangkamon and
             Nutanong, Sarana and Chuangsuwanich, Ekapol},
  journal = {arXiv preprint arXiv:2507.09618},
  year    = {2025}
}

@inproceedings{voxvietnam,
  title     = {VoxVietnam: a Large-Scale Multi-Genre Dataset for Vietnamese
               Speaker Recognition},
  author    = {Vu, Hoang Long and Dat, Phuong Tuan and Nhi, Pham Thao and
               Hao, Nguyen Song and Thu Trang, Nguyen Thi},
  booktitle = {ICASSP 2025 - 2025 IEEE International Conference on Acoustics,
               Speech and Signal Processing (ICASSP)},
  pages     = {1--5},
  year      = {2025},
  doi       = {10.1109/ICASSP49660.2025.10890124}
}
Downloads last month
82
Number of rows:
8,827
Total file size:
2.21 GB

Collection including MERaLiON/sea_audiobench_datasets_SpR

Paper for MERaLiON/sea_audiobench_datasets_SpR