---
dataset_info:
- config_name: amharic
  features:
  - name: audio
    dtype:
      audio:
        sampling_rate: 16000
        decode: false
  - name: transcript
    dtype: string
  - name: duration
    dtype: float32
  - name: language
    dtype: string
  - name: source_dataset
    dtype: string
  splits:
  - name: train
    num_bytes: 17364636006
    num_examples: 55056
  - name: validation
    num_bytes: 2168740969
    num_examples: 6882
  - name: test
    num_bytes: 2159563988
    num_examples: 6882
  download_size: 21731893985
  dataset_size: 21692940963
- config_name: arabic_msa
  features:
  - name: audio
    dtype:
      audio:
        sampling_rate: 16000
        decode: false
  - name: transcript
    dtype: string
  - name: duration
    dtype: float32
  - name: language
    dtype: string
  - name: source_dataset
    dtype: string
  splits:
  - name: train
    num_bytes: 13225400406
    num_examples: 101129
  - name: validation
    num_bytes: 1636700524
    num_examples: 12640
  - name: test
    num_bytes: 1665484265
    num_examples: 12640
  download_size: 16637043127
  dataset_size: 16527585195
- config_name: assamese
  features:
  - name: audio
    dtype:
      audio:
        sampling_rate: 16000
        decode: false
  - name: transcript
    dtype: string
  - name: duration
    dtype: float32
  - name: language
    dtype: string
  - name: source_dataset
    dtype: string
  splits:
  - name: train
    num_bytes: 18106046593
    num_examples: 90707
  - name: validation
    num_bytes: 2361874099
    num_examples: 11338
  - name: test
    num_bytes: 2212981640
    num_examples: 11338
  download_size: 23062364509
  dataset_size: 22680902332
- config_name: bengali
  features:
  - name: audio
    dtype:
      audio:
        sampling_rate: 16000
        decode: false
  - name: transcript
    dtype: string
  - name: duration
    dtype: float32
  - name: language
    dtype: string
  - name: source_dataset
    dtype: string
  splits:
  - name: train
    num_bytes: 18300068424
    num_examples: 129168
  - name: validation
    num_bytes: 2329082364
    num_examples: 16146
  - name: test
    num_bytes: 2319687239
    num_examples: 16146
  download_size: 23063782988
  dataset_size: 22948838027
- config_name: czech
  features:
  - name: audio
    dtype:
      audio:
        sampling_rate: 16000
        decode: false
  - name: transcript
    dtype: string
  - name: duration
    dtype: float32
  - name: language
    dtype: string
  - name: source_dataset
    dtype: string
  splits:
  - name: train
    num_bytes: 4841007906
    num_examples: 24823
  - name: validation
    num_bytes: 572497401
    num_examples: 3103
  download_size: 5194588398
  dataset_size: 5413505307
configs:
- config_name: amharic
  data_files:
  - split: train
    path: languages/Amharic/train-*
  - split: validation
    path: languages/Amharic/validation-*
  - split: test
    path: languages/Amharic/test-*
- config_name: arabic_msa
  data_files:
  - split: train
    path: languages/Arabic MSA/train-*
  - split: validation
    path: languages/Arabic MSA/validation-*
  - split: test
    path: languages/Arabic MSA/test-*
- config_name: assamese
  data_files:
  - split: train
    path: languages/Assamese/train-*
  - split: validation
    path: languages/Assamese/validation-*
  - split: test
    path: languages/Assamese/test-*
- config_name: bengali
  data_files:
  - split: train
    path: languages/Bengali/train-*
  - split: validation
    path: languages/Bengali/validation-*
  - split: test
    path: languages/Bengali/test-*
- config_name: czech
  data_files:
  - split: train
    path: languages/Czech/train-*
  - split: validation
    path: languages/Czech/validation-*
tags:
- audio
- speech-recognition
- automatic-speech-recognition
- multilingual
- amharic
- arabic-msa
- assamese
- bengali
- czech
- dutch
- egyptian-arabic
- english
- farsi-persian
- filipino-tagalog
- french
- german
- greek
- gujarati
- gulf-arabic
- hebrew
- hindi
- hinglish-hindi-english
- hausa
- indonesian
- italian
- japanese
- kannada
- kazakh
- khmer
- korean
- malay
- malayalam
- mandarin-chinese
- marathi
- moroccan-arabic
- nepali
- odia
- polish
- portuguese
- punjabi
- romanian
- russian
- sanskrit
- saudi-arabic
- spanish
- swahili
- tamil
- telugu
- thai
- tunisian-arabic
- turkish
- ukrainian
- urdu
- vietnamese
language:
- am
- ar
- as
- bn
- cs
- nl
- en
- fa
- fil
- fr
- de
- el
- gu
- he
- hi
- ha
- id
- it
- ja
- kn
- kk
- km
- ko
- ms
- ml
- zh
- mr
- ne
- or
- pl
- pt
- pa
- ro
- ru
- sa
- es
- sw
- ta
- te
- th
- tr
- uk
- ur
- vi
---
# Multilingual STT Dataset

Audio and transcript pairs for 50 languages. Each language is a Dataset Viewer configuration with train, validation, and test splits.

## Language configurations

- `amharic`: Amharic
- `arabic_msa`: Arabic MSA
- `assamese`: Assamese
- `bengali`: Bengali
- `czech`: Czech
- `dutch`: Dutch
- `egyptian_arabic`: Egyptian Arabic
- `english`: English
- `farsi_persian`: Farsi - Persian
- `filipino_tagalog`: Filipino - Tagalog
- `french`: French
- `german`: German
- `greek`: Greek
- `gujarati`: Gujarati
- `gulf_arabic`: Gulf Arabic
- `hebrew`: Hebrew
- `hindi`: Hindi
- `hinglish_hindi_english`: Hinglish - Hindi-English
- `huasa`: Huasa
- `indonesian`: Indonesian
- `italian`: Italian
- `japanese`: Japanese
- `kannada`: Kannada
- `kazakh`: Kazakh
- `khmer`: Khmer
- `korean`: Korean
- `malay`: Malay
- `malayalam`: Malayalam
- `mandarin_chinese`: Mandarin Chinese
- `marathi`: Marathi
- `moroccan_arabic`: Moroccan Arabic
- `nepali`: Nepali
- `odia`: Odia
- `polish`: Polish
- `portuguese`: Portuguese
- `punjabi`: Punjabi
- `romanian`: Romanian
- `russian`: Russian
- `sanskrit`: Sanskrit
- `saudi_arabic`: Saudi Arabic
- `spanish`: Spanish
- `swahili`: Swahili
- `tamil`: Tamil
- `telugu`: Telugu
- `thai`: Thai
- `tunisian_arabic`: Tunisian Arabic
- `turkish`: Turkish
- `ukrainian`: Ukrainian
- `urdu`: Urdu
- `vietnamese`: Vietnamese