metadata

dataset_info:
  features:
    - name: id
      dtype: string
    - name: meeting_id
      dtype: string
    - name: speaker_id
      dtype: string
    - name: start_time
      dtype: float64
    - name: end_time
      dtype: float64
    - name: duration
      dtype: float64
    - name: has_audio
      dtype: bool
    - name: has_video
      dtype: bool
    - name: has_lip_video
      dtype: bool
    - name: disfluency_type
      dtype: string
    - name: transcript
      dtype: string
    - name: audio
      dtype: string
    - name: video
      dtype: string
    - name: video_format
      struct:
        - name: fps
          dtype: float64
    - name: audio_format
      struct:
        - name: sampling_rate
          dtype: int64
    - name: lip_video
      dtype: string
    - name: lip_video_format
      struct:
        - name: fps
          dtype: float64
  splits:
    - name: train
      num_bytes: 8928921
      num_examples: 35731
  download_size: 3572005
  dataset_size: 8928921
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
task_categories:
  - automatic-speech-recognition
language:
  - en
tags:
  - audio-visual
  - paralinguistic
  - disfluencies
size_categories:
  - 10K<n<100K
pretty_name: AMI Disfluency Laughter Dataset

AMI DisfluencyLaughter Dataset

This dataset contains segmented audio and video clips which extract from AMI Meeting Corpus. The segmented audio/videos created in this dataset are mainly the disfluencies and laughter events, extracted from original recordings.

General information about this dataset:

Number of recordings: 35,731
Has audio: True
Has video: True
Has lip video: True

Dataset({
    features: ['id', 'meeting_id', 'speaker_id', 'start_time', 'end_time', 'duration', 'has_audio', 'has_video', 'has_lip_video', 'disfluency_type', 'transcript', 'audio', 'video', 'lip_video'],
    num_rows: 35731
})

Dataset Structure

The dataset contains the following metadata:

{
    "id" : "unique segment id of the recordings",
    "meeting_id" : "meeting id the clips corresponding",
    "speaker_id" : "speaker id the clips corresponding",
    "start_time": "start timestamp of the disfluency/laughter event",
    "end_time": "end timestamp  of  the disfluency/laughter event",
    "duration": "duration of the event",
    "has_audio": "",
    "has_video": "",
    "has_lip_video":"",
    "disfluency_type": "Type of disfluency based on AMI annotation, type=laugh if it is laughter",
    "transcript":"the transcribed word corresponding to this disfluency, <laugh> if it is laughter",
    "audio": "path to the audio file",
    "video": "path to the video file",
    "lip_video": "path to the lip video"
}