object_v / README.md
MatsRooth's picture
Update README.md
d95de4d verified
metadata
dataset_info:
  features:
    - name: audio
      dtype: audio
    - name: label
      dtype:
        class_label:
          names:
            '0': down
            '1': left
            '2': 'no'
            '3': object01
            '4': object10
            '5': 'off'
            '6': 'on'
            '7': present01
            '8': present10
            '9': right
            '10': stop
            '11': up
            '12': 'yes'
  splits:
    - name: train
      num_bytes: 339891481.276
      num_examples: 12316
    - name: validation
      num_bytes: 59778691.086
      num_examples: 2174
    - name: train_object10
      num_bytes: 26539323.751217928
      num_examples: 962
    - name: train_object01
      num_bytes: 26704900.712244235
      num_examples: 968
    - name: validation_object10
      num_bytes: 5112635.276908924
      num_examples: 186
    - name: validation_object01
      num_bytes: 4947727.622815087
      num_examples: 180
    - name: train_present10
      num_bytes: 27422165.543358233
      num_examples: 994
    - name: train_present01
      num_bytes: 17131956.4662228
      num_examples: 621
    - name: validation_present10
      num_bytes: 4233048.788408464
      num_examples: 154
    - name: validation_present01
      num_bytes: 2556321.138454462
      num_examples: 93
  download_size: 447818847
  dataset_size: 514318251.6616301
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
      - split: train_object10
        path: data/train_object10-*
      - split: train_object01
        path: data/train_object01-*
      - split: validation_object10
        path: data/validation_object10-*
      - split: validation_object01
        path: data/validation_object01-*
      - split: train_present10
        path: data/train_present10-*
      - split: train_present01
        path: data/train_present01-*
      - split: validation_present10
        path: data/validation_present10-*
      - split: validation_present01
        path: data/validation_present01-*

This is a draft dataset of prosodic minimal pairs. It includes two stress doublets, the noun and verb uses of object and of present.
Remaining words are from superb ks. object10 and present10 tokens lexically have left stress. object01 and present01 tokens lexically have right stress.

There are separate validation splits for each of object10, present10, object01 and present01. In each of these splits, all the tokens have the same segmental content and lexical prosody. Test splits are not here yet.