Natural-Text / README.md
qingy2019's picture
Upload dataset
e7c7250 verified
metadata
dataset_info:
  features:
    - name: transcript
      dtype: string
  splits:
    - name: train
      num_bytes: 142345568
      num_examples: 76241
  download_size: 86065531
  dataset_size: 142345568
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

This dataset consists of a mixture of TED talk transcripts, the high-quality Aeon essays, as well as a filtered selection of NPR Interview Transcripts.

All NPR transcripts were filtered to contain only the turns with > 100 words.