---
dataset_info:
  features:
  - name: transcript
    dtype: string
  splits:
  - name: train
    num_bytes: 165824830
    num_examples: 106241
  download_size: 101559053
  dataset_size: 165824830
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

This dataset consists of a mixture of [TED talk transcripts](https://www.kaggle.com/datasets/miguelcorraljr/ted-ultimate-dataset), the high-quality [Aeon essays](https://www.kaggle.com/datasets/mannacharya/aeon-essays-dataset), as well as a filtered selection of [NPR Interview Transcripts](https://www.kaggle.com/datasets/shuyangli94/interview-npr-media-dialog-transcripts).

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