Stay organized with collections Save and categorize content based on your preferences.

  • Description:

SAMSum Corpus contains over 16k chat dialogues with manually annotated summaries.

There are two features:

  • dialogue: text of dialogue.
  • summary: human written summary of the dialogue.
  • id: id of a example.

  • Homepage:

  • Source code: tfds.summarization.Samsum

  • Versions:

    • 1.0.0 (default): No release notes.
  • Download size: Unknown size

  • Dataset size: 10.71 MiB

  • Manual download instructions: This dataset requires you to download the source data manually into download_config.manual_dir (defaults to ~/tensorflow_datasets/downloads/manual/):
    Download, decompress and place train.json, val.json and test.json in the manual follder.

  • Auto-cached (documentation): Yes

  • Splits:

Split Examples
'test' 819
'train' 14,732
'validation' 818
  • Feature structure:
    'dialogue': Text(shape=(), dtype=tf.string),
    'id': Text(shape=(), dtype=tf.string),
    'summary': Text(shape=(), dtype=tf.string),
  • Feature documentation:
Feature Class Shape Dtype Description
dialogue Text tf.string
id Text tf.string
summary Text tf.string
  • Citation:
  title={SAMSum Corpus: A Human-annotated Dialogue Dataset for Abstractive Summarization},
  author={Gliwa, Bogdan and Mochol, Iwona and Biesek, Maciej and Wawer, Aleksander},
  journal={arXiv preprint arXiv:1911.12237},