README.md

metadata

language:
  - en
license: apache-2.0
task_categories:
  - text-classification
task_ids:
  - natural-language-inference
dataset_info:
  - config_name: atomic
    features:
      - name: Premise
        dtype: string
      - name: Hypothesis
        dtype: string
      - name: Update
        dtype: string
      - name: UpdateType
        dtype: string
    splits:
      - name: train
        num_bytes: 4463463
        num_examples: 35002
      - name: validation
        num_bytes: 487230
        num_examples: 3840
      - name: test
        num_bytes: 526381
        num_examples: 4138
    download_size: 1871664
    dataset_size: 5477074
  - config_name: snli
    features:
      - name: Premise
        dtype: string
      - name: Hypothesis
        dtype: string
      - name: Update
        dtype: string
      - name: UpdateType
        dtype: string
    splits:
      - name: train
        num_bytes: 15781997
        num_examples: 88674
      - name: validation
        num_bytes: 328847
        num_examples: 1785
      - name: test
        num_bytes: 330956
        num_examples: 1837
    download_size: 3627986
    dataset_size: 16441800
  - config_name: social
    features:
      - name: Hypothesis
        dtype: string
      - name: Update
        dtype: string
      - name: UpdateType
        dtype: string
    splits:
      - name: train
        num_bytes: 9099604
        num_examples: 77016
      - name: validation
        num_bytes: 1094305
        num_examples: 9343
      - name: test
        num_bytes: 1113280
        num_examples: 9439
    download_size: 2959808
    dataset_size: 11307189
configs:
  - config_name: atomic
    data_files:
      - split: train
        path: atomic/train-*
      - split: validation
        path: atomic/validation-*
      - split: test
        path: atomic/test-*
  - config_name: snli
    data_files:
      - split: train
        path: snli/train-*
      - split: validation
        path: snli/validation-*
      - split: test
        path: snli/test-*
  - config_name: social
    data_files:
      - split: train
        path: social/train-*
      - split: validation
        path: social/validation-*
      - split: test
        path: social/test-*

https://github.com/rudinger/defeasible-nli

@inproceedings{rudinger2020thinking,
  title={Thinking like a skeptic: 
  feasible inference in natural language},
  author={Rudinger, Rachel and Shwartz, Vered and Hwang, Jena D and Bhagavatula, Chandra and Forbes, Maxwell and Le Bras, Ronan and Smith, Noah A and Choi, Yejin},
  booktitle={Findings of the Association for Computational Linguistics: EMNLP 2020},
  pages={4661--4675},
  year={2020}
}