PAC

This repository holds the PAC dataset for the EMNLP 2023 paper Crossing the Aisle: Unveiling Partisan and Counter-Partisan Events in News Reporting.

Contents

• data/annotations/: final annotations of each story.
• data/train.json: the training split that contains news articles from 2012 - 2021.
• data/dev.json: the development split that contains news articles from 2022 - 2022.06
• data/test.json: the test split that contains news articles from 2022.07 - 2023

The annotation file in data/annotations/ is formatted as follows:

{
  "title":
  "id":
  "split":
  "date":
  "topic":
  "source":
  "media_ideology":
  "relative_ideology":
  "absolute_ideology":
  "news":[{"id":, "text"}...]
  "events":[{"e_id":, "token":, "sentence_id":,"start":, "end":,"remove":,"partisanship":, "entity":}...]
  "entity_list":[{"text":, "ideology":, "id":, "events":[]}...]
}

source is the media outlet of the article.

media_ieology is the ideology of the media outlet.

remove is the field that indicates whether the event is removed by annotators.

The data JSON file in data is formatted as follows:

{
  "story_id": [article_1, article_2],...
}

Updates

• (10/31): Initial Data Release
• (02/07): Release individual annotations and merge entity annotations
• In Progress: Revise any found mistakes.

Citation

Please cite our paper if you find our dataset useful:

@inproceedings{zou-etal-2023-crossing,
    title = "Crossing the Aisle: Unveiling Partisan and Counter-Partisan Events in News Reporting",
    author = "Zou, Kaijian
      Zhang, Xinliang Frederick and
      Wu, Winston and
      Beauchamp, Nicholas  and
      Wang, Lu",
    booktitle = "Findings of the 2023 Conference on Empirical Methods in Natural Language Processing, {EMNLP} 2023",
    year = "2023",
    publisher = "Association for Computational Linguistics",
}