AbedGarra/ccnews2024plus

📰 CCNews-Plus: A Continuation of Stanford's CCNews Dataset

We continue the work started by stanford-oval/ccnews by extending the Common Crawl News corpus beyond June 2024, applying similar high-quality filtering and deduplication inspired by the datatrove methodology.

If you found our dataset useful, please cite our paper:

@article{fesenko2026forecasting,
         title = {Forecasting Cybersecurity Incidents Using Geopolitical Data and Large Language Models},
         author = {Fesenko, Mark and Garra, Abdullah and Harel, Yaniv and Sharif, Mahmood},
         eprint = {2610.04798},
         journal = {arXiv},
         year = {2026},
}

📦 Structure

• Language: English only (language: "en")
• Each monthly split is provided in its own subdirectory (2024_06/, 2024_07/, etc.)
• Inside each folder: .jsonl.gz files, where each line is a deduplicated, filtered news article
• June 2024 overlap: We include June 2024 for comparison with Stanford's original dataset

Data Schema

Each article contains 3 main fields:

• text: The full article content (extracted via trafilatura)
• _id: Unique identifier for the article
• metadata: Rich metadata object containing:

{
  "url": "https://www.wptv.com/sports/mike-tyson-vs-jake-paul-boxing-match-postponed-due-to-tysons-medical-emergency",
  "date": "2024-05-31T00:00:00",           // Article publication date (trafilatura)
  "warc_date": "2024-06-01T00:16:20",      // When crawled by Common Crawl
  "publisher": "wptv.com",                 // Publisher domain
  "title": "Mike Tyson vs. Jake Paul boxing match postponed due to Tyson's medical emergency",
  "author": "Alex Arger",                  // Article author
  "tags": [                                // Article tags
    "Scripps News",
    "Sports", 
    "Celebrity"
  ],
  "categories": ["Sports"],                // Article categories
  "sitename": "WPTV News Channel 5 West Palm",
  "file_path": "s3://commoncrawl/crawl-data/CC-NEWS/2024/06/CC-NEWS-20240601001620-04355.warc.gz",
  "language": "en",                        // Language detection
  "language_score": 0.9763813018798828,   // Language confidence score  
  "token_count": 699                       // Number of tokens in article
}

The extra metadata fields are extracted using trafilatura, with date representing the actual publication date of the article, falling back to the crawl date if failed.

📥 Loading Example

from datasets import load_dataset

# 1st Use Case: 
# Load specific months
dataset_june = load_dataset("AbedGarra/ccnews2024plus", "2024_06")
dataset_july = load_dataset("AbedGarra/ccnews2024plus", "2024_07")
dataset_august = load_dataset("AbedGarra/ccnews2024plus", "2024_08")


# 2nd Use Case:
# Stream and preview first 10 articles
dataset = load_dataset("AbedGarra/ccnews2024plus", "2024_07", streaming=True)
for i, article in enumerate(dataset['train']):
    if i >= 10:
        break
    print(f"Article {i+1}:")
    print(f"Title: {article['metadata']['title']}")
    print(f"Publisher: {article['metadata']['publisher']}")
    print(f"Date: {article['metadata']['date']}")
    print(f"Text preview: {article['text'][:200]}...")
    print("-" * 50)

Available Subsets

• Articles from 2024_06 to 2026_04.