README.md

metadata

dataset_info:
  - config_name: ASR
    features:
      - name: id
        dtype: string
      - name: audio
        dtype: audio
      - name: sentence
        dtype: string
      - name: metadata
        struct:
          - name: audio
            struct:
              - name: recordedID
                dtype: string
              - name: recordedDate
                dtype: string
              - name: recordedStart
                dtype: string
              - name: fileName
                dtype: string
              - name: filePath
                dtype: string
              - name: recordedTime
                dtype: float32
              - name: recordedQuality
                dtype: string
              - name: recordedDevice
                dtype: string
              - name: recordedDeviceName
                dtype: string
              - name: recordedEnv
                dtype: string
          - name: recordPerson
            struct:
              - name: collectedID
                dtype: string
              - name: recorderID
                dtype: string
              - name: sex
                dtype: string
              - name: generation
                dtype: string
              - name: residence
                dtype: string
              - name: dialect
                dtype: string
              - name: dialectRegion
                dtype: string
              - name: areaInfomation
                struct:
                  - name: fatherHometown
                    dtype: string
                  - name: motherHometown
                    dtype: string
    splits:
      - name: train
        num_bytes: 1019163173828.5
        num_examples: 3621732
      - name: validation
        num_bytes: 121672797307.125
        num_examples: 429055
    download_size: 1131329397132
    dataset_size: 1140835971135.625
  - config_name: Original
    features:
      - name: audio
        struct:
          - name: recordedID
            dtype: string
          - name: recordedDate
            dtype: string
          - name: recordedStart
            dtype: string
          - name: fileName
            dtype: string
          - name: filePath
            dtype: string
          - name: recordedTime
            dtype: float32
          - name: recordedQuality
            dtype: string
          - name: recordedDevice
            dtype: string
          - name: recordedDeviceName
            dtype: string
          - name: recordedEnv
            dtype: string
          - name: audio
            dtype: audio
      - name: script
        struct:
          - name: scriptId
            dtype: string
          - name: scriptITN
            dtype: string
          - name: scriptTN
            dtype: string
          - name: scriptNumberWord
            dtype: string
          - name: patternTheme
            dtype: string
      - name: recordPerson
        struct:
          - name: collectedID
            dtype: string
          - name: recorderID
            dtype: string
          - name: sex
            dtype: string
          - name: generation
            dtype: string
          - name: residence
            dtype: string
          - name: dialect
            dtype: string
          - name: dialectRegion
            dtype: string
          - name: areaInfomation
            struct:
              - name: fatherHometown
                dtype: string
              - name: motherHometown
                dtype: string
    splits:
      - name: train
        num_bytes: 1173145614684.125
        num_examples: 4137031
      - name: validation
        num_bytes: 139858660796.125
        num_examples: 489607
    download_size: 1301404928518
    dataset_size: 1313004275480.25
configs:
  - config_name: ASR
    data_files:
      - split: train
        path: ASR/train-*
      - split: validation
        path: ASR/validation-*
  - config_name: Original
    data_files:
      - split: train
        path: Original/train-*
      - split: validation
        path: Original/validation-*
task_categories:
  - automatic-speech-recognition
language:
  - ko
size_categories:
  - n>1T

jp1924/TNT0.5B-Dual 모델을 활용해서 이중 전사문으로 변환한 데이터