README.md

metadata

dataset_info:
  features:
    - name: premise
      dtype: string
    - name: hypothesis
      dtype: string
    - name: task_name
      dtype: string
    - name: label_name
      dtype: string
  splits:
    - name: train
      num_bytes: 317612708
      num_examples: 1018733
    - name: test
      num_bytes: 15622840
      num_examples: 59140
  download_size: 212682398
  dataset_size: 333235548
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*

NLI Mix Zero-Shot

This dataset is a single dataset entry point for the following train and test datasets:

• train: MoritzLaurer/dataset_train_nli
• test: MoritzLaurer/dataset_test_concat_nli

Datasets consists of a mixture of text classification datasets using the NLI (Natural Language Inference) format. It can be use ot train a powerful Zero-Shot Text Classification (ZS-TC) model.

For more details on the creation of the dataset (datasets used, datasets format, datasets cleaning, ...) please refer to the page of each dataset.

All the credits goes to MoritzLaurer. Thank you for your hard work and sharing it with the community!