Multilingual Speech Recognition Evaluation Data Set 2
  (SPREDS 2: SPeech Recognition Evaluation Data Set 2)

    Overview: This data set is evaluation data for multilingual speech recognition. It consists of speech and transcriptions recorded under almost the same conditions as domain, number of people, recording environment etc.

    Directory structure: Directory structure after decompressing xz, tar
    ・Complete set(10lang_all)
    $ver/
     $lang/
      LABEL/
       SPREDS2.$ver.$lang.label
       SPREDS2.$ver.$lang.info
      WAVE/
       *.wav
      00README.txt
     doc/
      GCP_DialectCode_v1.0.3.xlsx

    ・Each language
    $lang/
     LABEL/
      SPREDS2.$ver.$lang.label
      SPREDS2.$ver.$lang.info
     WAVE/
      *.wav
     00README.txt

    $ver =[version number]
    $lang={ja,en,zh,ko,th,vi,id,my,es,fr}

    License | Download