README.md

metadata

dataset_info:
  features:
    - name: topic
      dtype: string
    - name: source
      dtype: string
    - name: bias
      dtype:
        class_label:
          names:
            '0': '0'
            '1': '1'
            '2': '2'
    - name: url
      dtype: string
    - name: title
      dtype: string
    - name: date
      dtype: string
    - name: authors
      dtype: string
    - name: content
      dtype: string
    - name: content_original
      dtype: string
    - name: source_url
      dtype: string
    - name: bias_text
      dtype:
        class_label:
          names:
            '0': left
            '1': right
            '2': center
    - name: ID
      dtype: string
  splits:
    - name: train
      num_bytes: 342169887
      num_examples: 27978
    - name: test
      num_bytes: 15232421
      num_examples: 1300
    - name: valid
      num_bytes: 84576118
      num_examples: 6996
  download_size: 262975174
  dataset_size: 441978426
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*
      - split: valid
        path: data/valid-*
license: apache-2.0
language:
  - en
pretty_name: News Articles with Political Bias Annotations

News Articles with Political Bias Annotations (All Articles)

Source

Derived from Baly et al.'s work: We Can Detect Your Bias: Predicting the Political Ideology of News Articles (Baly et al., EMNLP 2020)

Information

This dataset contains 37,554 news articles manually annotated for political ideology, either "left", "center", or "right". This version contains randomly selected test/training/validation splits, identical to those used by Baly (according to their git repository).

In the Baly dataset, there are 1280 unsorted articles that are not associated with the train, test, or validation sets. These 1280 articles have been appended to the training set. To access the dataset with exactly Baly's splits, see siddharthmb/article-bias-prediction-random-splits.