opinion_abstracts

  • Description:

There are two sub datasets:

(1) RottenTomatoes: The movie critics and consensus crawled from http://rottentomatoes.com/ It has fields of "_movie_name", "_movie_id", "_critics", and "_critic_consensus".

(2) IDebate: The arguments crawled from http://idebate.org/ It has fields of "_debate_name", "_debate_id", "_claim", "_claim_id", "_argument_sentences".

@inproceedings{wang-ling-2016-neural,
    title = "Neural Network-Based Abstract Generation for Opinions and Arguments",
    author = "Wang, Lu  and
      Ling, Wang",
    booktitle = "Proceedings of the 2016 Conference of the North {A}merican Chapter of the Association for Computational Linguistics: Human Language Technologies",
    month = jun,
    year = "2016",
    address = "San Diego, California",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/N16-1007",
    doi = "10.18653/v1/N16-1007",
    pages = "47--57",
}

opinion_abstracts/rotten_tomatoes (default config)

  • Config description: Professional critics and consensus of 3,731 movies.

  • Dataset size: 50.10 MiB

  • Splits:

Split Examples
'train' 3,731
  • Features:
FeaturesDict({
    '_critic_consensus': tf.string,
    '_critics': Sequence({
        'key': tf.string,
        'value': tf.string,
    }),
    '_movie_id': tf.string,
    '_movie_name': tf.string,
})

opinion_abstracts/idebate

  • Config description: 2,259 claims for 676 debates.

  • Dataset size: 3.15 MiB

  • Splits:

Split Examples
'train' 2,259
  • Features:
FeaturesDict({
    '_argument_sentences': Sequence({
        'key': tf.string,
        'value': tf.string,
    }),
    '_claim': tf.string,
    '_claim_id': tf.string,
    '_debate_name': tf.string,
})