README.md

metadata

dataset_info:
  features:
    - name: RawDataInfo
      struct:
        - name: RawDataId
          dtype: string
        - name: Copyrighter
          dtype: string
        - name: SampleRate(Hz)
          dtype: int32
        - name: Channel
          dtype: int32
        - name: BitDepth(bit)
          dtype: int32
        - name: RecordingDevice
          dtype: string
        - name: BitRate(kbps)
          dtype: int32
        - name: CollectionType
          dtype: string
        - name: RecDateTime
          dtype: string
        - name: RecDataLength(sec)
          dtype: int32
        - name: Season
          dtype: string
        - name: Weather
          dtype: string
        - name: TimeZone
          dtype: string
        - name: PlaceType
          dtype: string
        - name: DistanceType
          dtype: string
        - name: FileExtension
          dtype: string
    - name: SourceDataInfo
      struct:
        - name: SourceDataId
          dtype: string
        - name: FileExtension
          dtype: string
        - name: NoOfClip
          dtype: int32
        - name: ClipDataLength(sec)
          dtype: int32
    - name: LabelDataInfo
      struct:
        - name: Path
          dtype: string
        - name: LabelID
          dtype: string
        - name: NumAnnotator
          dtype: int32
        - name: Division1
          dtype: string
        - name: Division2
          dtype: string
        - name: Class
          dtype: string
        - name: Desc
          dtype: string
        - name: Type
          dtype: string
        - name: NumSegmentation
          dtype: int32
        - name: Segmentations
          list:
            sequence: float32
    - name: audio
      dtype:
        audio:
          sampling_rate: 44100
  splits:
    - name: train
      num_bytes: 20358466791
      num_examples: 35848
    - name: validation
      num_bytes: 2273622062.875
      num_examples: 4481
  download_size: 20620542669
  dataset_size: 22632088853.875
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
task_categories:
  - automatic-speech-recognition
language:
  - ko
tags:
  - STT
  - Audio
  - Noise
size_categories:
  - 10B<n<100B

노이즈 데이터인데 노이즈 앞쪽에 해당 음성은 ~~해서 어쩌구 저쩌구 들어가 있음. 전처리 할 때 참고