Hugging Face's logo

Datasets:
disco-eth
/
WorldSpeech

Tasks:
Automatic Speech Recognition
Text-to-Speech
Audio Classification
Modalities:
Audio
Languages:
Afrikaans
Amharic
Arabic
Size:
10M<n<100M
ArXiv:
Tags:
speech
multilingual
low-resource
parliamentary
asr
tts
DOI:
License:
Dataset card Files
xet
 Community
1
WorldSpeech / README.md
AAdonis's picture
AAdonis
Fill es_uy: 894.4h / 2.94 OVR
4e7a13c verified
preview code
|
raw
history blame
18.5 kB
metadata 
license: other
multilinguality:
  - multilingual
viewer: true
configs:
  - config_name: af_za
    data_files:
      - split: test
        path: data/af_za/test-*.parquet
      - split: train
        path: data/af_za/train-*.parquet
  - config_name: am_et
    data_files:
      - split: test
        path: data/am_et/test-*.parquet
      - split: train
        path: data/am_et/train-*.parquet
  - config_name: ar_bh
    data_files:
      - split: test
        path: data/ar_bh/test-*.parquet
      - split: train
        path: data/ar_bh/train-*.parquet
  - config_name: ar_dz
    data_files:
      - split: test
        path: data/ar_dz/test-*.parquet
      - split: train
        path: data/ar_dz/train-*.parquet
  - config_name: ar_eg
    data_files:
      - split: test
        path: data/ar_eg/test-*.parquet
      - split: train
        path: data/ar_eg/train-*.parquet
  - config_name: ar_iq
    data_files:
      - split: test
        path: data/ar_iq/test-*.parquet
      - split: train
        path: data/ar_iq/train-*.parquet
  - config_name: ar_kw
    data_files:
      - split: test
        path: data/ar_kw/test-*.parquet
      - split: train
        path: data/ar_kw/train-*.parquet
  - config_name: ar_ma
    data_files:
      - split: test
        path: data/ar_ma/test-*.parquet
      - split: train
        path: data/ar_ma/train-*.parquet
  - config_name: ar_sa
    data_files:
      - split: test
        path: data/ar_sa/test-*.parquet
      - split: train
        path: data/ar_sa/train-*.parquet
  - config_name: ar_tn
    data_files:
      - split: test
        path: data/ar_tn/test-*.parquet
      - split: train
        path: data/ar_tn/train-*.parquet
  - config_name: ar_un
    data_files:
      - split: test
        path: data/ar_un/test-*.parquet
      - split: train
        path: data/ar_un/train-*.parquet
  - config_name: az_az
    data_files:
      - split: test
        path: data/az_az/test-*.parquet
      - split: train
        path: data/az_az/train-*.parquet
  - config_name: be_by
    data_files:
      - split: test
        path: data/be_by/test-*.parquet
      - split: train
        path: data/be_by/train-*.parquet
  - config_name: bn_bd
    data_files:
      - split: test
        path: data/bn_bd/test-*.parquet
      - split: train
        path: data/bn_bd/train-*.parquet
  - config_name: ckb_iq
    data_files:
      - split: test
        path: data/ckb_iq/test-*.parquet
      - split: train
        path: data/ckb_iq/train-*.parquet
  - config_name: cnr_me
    data_files:
      - split: test
        path: data/cnr_me/test-*.parquet
      - split: train
        path: data/cnr_me/train-*.parquet
  - config_name: crs_sc
    data_files:
      - split: test
        path: data/crs_sc/test-*.parquet
      - split: train
        path: data/crs_sc/train-*.parquet
  - config_name: cs_cz
    data_files:
      - split: test
        path: data/cs_cz/test-*.parquet
      - split: train
        path: data/cs_cz/train-*.parquet
  - config_name: de_at
    data_files:
      - split: test
        path: data/de_at/test-*.parquet
      - split: train
        path: data/de_at/train-*.parquet
  - config_name: dv_mv
    data_files:
      - split: test
        path: data/dv_mv/test-*.parquet
      - split: train
        path: data/dv_mv/train-*.parquet
  - config_name: el_cy
    data_files:
      - split: test
        path: data/el_cy/test-*.parquet
      - split: train
        path: data/el_cy/train-*.parquet
  - config_name: el_gr
    data_files:
      - split: test
        path: data/el_gr/test-*.parquet
      - split: train
        path: data/el_gr/train-*.parquet
  - config_name: en_au
    data_files:
      - split: test
        path: data/en_au/test-*.parquet
      - split: train
        path: data/en_au/train-*.parquet
  - config_name: en_jm
    data_files:
      - split: test
        path: data/en_jm/test-*.parquet
      - split: train
        path: data/en_jm/train-*.parquet
  - config_name: en_ke
    data_files:
      - split: test
        path: data/en_ke/test-*.parquet
      - split: train
        path: data/en_ke/train-*.parquet
  - config_name: en_nz
    data_files:
      - split: test
        path: data/en_nz/test-*.parquet
      - split: train
        path: data/en_nz/train-*.parquet
  - config_name: en_sl
    data_files:
      - split: test
        path: data/en_sl/test-*.parquet
      - split: train
        path: data/en_sl/train-*.parquet
  - config_name: en_zm
    data_files:
      - split: test
        path: data/en_zm/test-*.parquet
      - split: train
        path: data/en_zm/train-*.parquet
  - config_name: eo
    data_files:
      - split: test
        path: data/eo/test-*.parquet
      - split: train
        path: data/eo/train-*.parquet
  - config_name: es_ar
    data_files:
      - split: test
        path: data/es_ar/test-*.parquet
      - split: train
        path: data/es_ar/train-*.parquet
  - config_name: es_cl
    data_files:
      - split: test
        path: data/es_cl/test-*.parquet
      - split: train
        path: data/es_cl/train-*.parquet
  - config_name: es_co
    data_files:
      - split: test
        path: data/es_co/test-*.parquet
      - split: train
        path: data/es_co/train-*.parquet
  - config_name: es_es
    data_files:
      - split: train
        path: data/es_es/train-*.parquet
  - config_name: es_mx
    data_files:
      - split: test
        path: data/es_mx/test-*.parquet
      - split: train
        path: data/es_mx/train-*.parquet
  - config_name: es_pe
    data_files:
      - split: test
        path: data/es_pe/test-*.parquet
      - split: train
        path: data/es_pe/train-*.parquet
  - config_name: es_pr
    data_files:
      - split: test
        path: data/es_pr/test-*.parquet
      - split: train
        path: data/es_pr/train-*.parquet
  - config_name: es_py
    data_files:
      - split: test
        path: data/es_py/test-*.parquet
      - split: train
        path: data/es_py/train-*.parquet
  - config_name: es_uy
    data_files:
      - split: test
        path: data/es_uy/test-*.parquet
      - split: train
        path: data/es_uy/train-*.parquet
  - config_name: nl_nl
    data_files:
      - split: test
        path: data/nl_nl/test-*.parquet
      - split: train
        path: data/nl_nl/train-*.parquet

WorldSpeech

Highly multilingual, large-scale paired speech corpus assembled from primary public archives such as national parliaments, public broadcasters, and public-domain audiobooks across 107 language-country configs, with strong non-European and low-resource coverage.

Every segment ships with a verified human transcript (not ASR pseudo-labels), an ASR transcript, the CER between them, a WADA-SNR estimate, and four DNSMOS-P.835 quality scores, so you can filter, rank, or stratify training data without rerunning a single signal-quality model.

Audio is OGG Opus 24 kHz mono ~64 kbps VBR. Splits are 95/5, full provenance (source, source_url, session_date, source_start_s/source_end_s, segment_id) is preserved on every row, so you can trace any clip back to its original parliamentary record or broadcast.

Loading 

from datasets import load_dataset, Audio

ds = load_dataset("AAdonis/WorldSpeech", "nl_nl", split="train")
ds = ds.cast_column("audio", Audio(sampling_rate=24000))  # decodes Opus -> {"array": np.ndarray, "sampling_rate": 24000}

row = ds[0]
wav = row["audio"]["array"]            # 1-D float numpy array, 24 kHz mono
sr  = row["audio"]["sampling_rate"]    # 24000
text   = row["human_transcript"]
snr    = row["snr"]
dnsmos = row["dnsmos_ovr"]
print(len(wav)/sr, "sec", text[:60], snr, dnsmos)

# Streaming (no full download):
ds = load_dataset("AAdonis/WorldSpeech", "nl_nl", split="train", streaming=True)
ds = ds.cast_column("audio", Audio(sampling_rate=24000))
for row in ds:
    wav = row["audio"]["array"]
    break

Configs

Config Country Language Hours Avg DNSMOS OVR
af_za South Africa Afrikaans 20.2 2.79
am_et Ethiopia Amharic 39.6 3.11
ar_bh Bahrain Arabic 219.5 2.72
ar_dz Algeria Arabic 92.9 2.66
ar_eg Egypt Arabic 22.0 2.86
ar_iq Iraq Arabic 246.0 2.94
ar_kw Kuwait Arabic 175.5 2.82
ar_ma Morocco Arabic 78.3 2.68
ar_sa Saudi Arabia Arabic 6.1 2.01
ar_tn Tunisia Arabic 50.9 2.74
ar_un United Nations Arabic 11.1 3.03
az_az Azerbaijan Azerbaijani 305.4 2.76
be_by Belarus Belarusian 24.2 3.10
bn_bd Bangladesh Bengali 46.1 3.03
ckb_iq Iraq Sorani Kurdish 35.3 2.93
cnr_me Montenegro Montenegrin 47.9 2.98
crs_sc Seychelles Kreol Seselwa 1572.3 3.15
cs_cz Czech Republic Czech 931.6 2.87
de_at Austria German 1077.5 2.93
dv_mv Maldives Dhivehi 20.0 2.92
el_cy Cyprus Greek 394.5 2.90
el_gr Greece Greek 35.9 3.37
en_au Australia English 568.4 3.02
en_jm Jamaica English 9.6 2.52
en_ke Kenya English 170.0 2.97
en_nz New Zealand English 438.6 2.68
en_sl Sierra Leone English 102.4 2.70
en_zm Zambia English 291.6 2.70
eo (international) Esperanto 15.0 3.27
es_ar Argentina Spanish 251.7 2.78
es_cl Chile Spanish 774.1 TBD
es_co Colombia Spanish 156.8 2.75
es_es Spain Spanish 510.5 2.52
es_mx Mexico Spanish 713.8 2.63
es_pe Peru Spanish 300.4 2.63
es_pr Puerto Rico Spanish 228.6 2.73
es_py Paraguay Spanish 133.2 3.14
es_uy Uruguay Spanish 894.4 2.94
fa_ir Iran Persian TBD TBD
fr_ca Canada French TBD TBD
fr_cd DR Congo French TBD TBD
fr_ci Cote d'Ivoire French TBD TBD
ga_ie Ireland Irish TBD TBD
grc_gr Greece Ancient Greek TBD TBD
ha_ng Nigeria Hausa TBD TBD
he_il Israel Hebrew TBD TBD
hi_in India Hindi TBD TBD
hu_hu Hungary Hungarian TBD TBD
hy_am Armenia Armenian TBD TBD
id_id Indonesia Indonesian TBD TBD
ig_ng Nigeria Igbo TBD TBD
iu_ca Canada Inuktitut TBD TBD
ja_jp Japan Japanese TBD TBD
ka_ge Georgia Georgian TBD TBD
kk_kz Kazakhstan Kazakh TBD TBD
km_kh Cambodia Khmer TBD TBD
ko_kr South Korea Korean TBD TBD
la_va Vatican Latin TBD TBD
lb_lu Luxembourg Luxembourgish TBD TBD
lo_la Laos Lao TBD TBD
mfe_mu Mauritius Mauritian Creole TBD TBD
mi_nz New Zealand Maori TBD TBD
ml_in India Malayalam TBD TBD
mn_mn Mongolia Mongolian TBD TBD
mr_in India Marathi TBD TBD
ms_my Malaysia Malay TBD TBD
my_mm Myanmar Burmese TBD TBD
ne_np Nepal Nepali TBD TBD
nl_be Belgium Dutch TBD TBD
nl_nl Netherlands Dutch 997.1 3.07
nr_za South Africa Southern Ndebele TBD TBD
nso_za South Africa Northern Sotho TBD TBD
om_et Ethiopia Oromo TBD TBD
pa_in India Punjabi TBD TBD
pl_pl Poland Polish TBD TBD
pt_br Brazil Portuguese TBD TBD
rm_ch Switzerland Romansh TBD TBD
ro_md Moldova Romanian TBD TBD
ro_ro Romania Romanian TBD TBD
ru_by Belarus Russian TBD TBD
ru_ru Russia Russian TBD TBD
rw_rw Rwanda Kinyarwanda TBD TBD
si_lk Sri Lanka Sinhala TBD TBD
sm_ws Samoa Samoan TBD TBD
sn_zw Zimbabwe Shona TBD TBD
sq_al Albania Albanian TBD TBD
sq_xk Kosovo Albanian TBD TBD
ss_za South Africa Swati TBD TBD
st_za South Africa Sesotho TBD TBD
sv_ax Aland Islands Swedish TBD TBD
sw_tz Tanzania Swahili TBD TBD
ta_lk Sri Lanka Tamil TBD TBD
th_th Thailand Thai TBD TBD
ti_et Ethiopia Tigrinya TBD TBD
tl_ph Philippines Tagalog TBD TBD
tn_bw Botswana Tswana TBD TBD
tn_za South Africa Tswana TBD TBD
tr_tr Turkey Turkish TBD TBD
ts_za South Africa Tsonga TBD TBD
ug_cn China Uyghur TBD TBD
uz_uz Uzbekistan Uzbek TBD TBD
ve_za South Africa Venda TBD TBD
vi_vn Vietnam Vietnamese TBD TBD
xh_za South Africa Xhosa TBD TBD
yue_hk Hong Kong Cantonese TBD TBD
zh_tw Taiwan Chinese (Mandarin) TBD TBD
zu_za South Africa Zulu TBD TBD
Total - - 12009.0 2.85

Licenses

Config Country / Language License
af_za South Africa / Afrikaans License
am_et Ethiopia / Amharic License
ar_bh Bahrain / Arabic License
ar_dz Algeria / Arabic License
ar_eg Egypt / Arabic License
ar_iq Iraq / Arabic License
ar_kw Kuwait / Arabic License
ar_ma Morocco / Arabic License
ar_sa Saudi Arabia / Arabic License
ar_tn Tunisia / Arabic License
ar_un United Nations / Arabic License
az_az Azerbaijan / Azerbaijani License
be_by Belarus / Belarusian License
bn_bd Bangladesh / Bengali License
ckb_iq Iraq / Sorani Kurdish License
cnr_me Montenegro / Montenegrin License
crs_sc Seychelles / Kreol Seselwa License
cs_cz Czech Republic / Czech License
de_at Austria / German License
dv_mv Maldives / Dhivehi License
el_cy Cyprus / Greek License
el_gr Greece / Greek License
en_au Australia / English License
en_jm Jamaica / English License
en_ke Kenya / English License
en_nz New Zealand / English License
en_sl Sierra Leone / English License
en_zm Zambia / English License
eo (international) / Esperanto License
es_ar Argentina / Spanish License
es_cl Chile / Spanish License
es_co Colombia / Spanish License
es_es Spain / Spanish License
es_mx Mexico / Spanish License
es_pe Peru / Spanish License
es_pr Puerto Rico / Spanish License
es_py Paraguay / Spanish License
es_uy Uruguay / Spanish License
fa_ir Iran / Persian License
fr_ca Canada / French License
fr_cd DR Congo / French License
fr_ci Cote d'Ivoire / French License
ga_ie Ireland / Irish License
grc_gr Greece / Ancient Greek License
ha_ng Nigeria / Hausa License
he_il Israel / Hebrew License
hi_in India / Hindi License
hu_hu Hungary / Hungarian License
hy_am Armenia / Armenian License
id_id Indonesia / Indonesian License
ig_ng Nigeria / Igbo License
iu_ca Canada / Inuktitut License
ja_jp Japan / Japanese License
ka_ge Georgia / Georgian License
kk_kz Kazakhstan / Kazakh License
km_kh Cambodia / Khmer License
ko_kr South Korea / Korean License
la_va Vatican / Latin License
lb_lu Luxembourg / Luxembourgish License
lo_la Laos / Lao License
mfe_mu Mauritius / Mauritian Creole License
mi_nz New Zealand / Maori License
ml_in India / Malayalam License
mn_mn Mongolia / Mongolian License
mr_in India / Marathi License
ms_my Malaysia / Malay License
my_mm Myanmar / Burmese License
ne_np Nepal / Nepali License
nl_be Belgium / Dutch License
nl_nl Netherlands / Dutch License
nr_za South Africa / Southern Ndebele License
nso_za South Africa / Northern Sotho License
om_et Ethiopia / Oromo License
pa_in India / Punjabi License
pl_pl Poland / Polish License
pt_br Brazil / Portuguese License
rm_ch Switzerland / Romansh License
ro_md Moldova / Romanian License
ro_ro Romania / Romanian License
ru_by Belarus / Russian License
ru_ru Russia / Russian License
rw_rw Rwanda / Kinyarwanda License
si_lk Sri Lanka / Sinhala License
sm_ws Samoa / Samoan License
sn_zw Zimbabwe / Shona License
sq_al Albania / Albanian License
sq_xk Kosovo / Albanian License
ss_za South Africa / Swati License
st_za South Africa / Sesotho License
sv_ax Aland Islands / Swedish License
sw_tz Tanzania / Swahili License
ta_lk Sri Lanka / Tamil License
th_th Thailand / Thai License
ti_et Ethiopia / Tigrinya License
tl_ph Philippines / Tagalog License
tn_bw Botswana / Tswana License
tn_za South Africa / Tswana License
tr_tr Turkey / Turkish License
ts_za South Africa / Tsonga License
ug_cn China / Uyghur License
uz_uz Uzbekistan / Uzbek License
ve_za South Africa / Venda License
vi_vn Vietnam / Vietnamese License
xh_za South Africa / Xhosa License
yue_hk Hong Kong / Cantonese License
zh_tw Taiwan / Chinese (Mandarin) License
zu_za South Africa / Zulu License