Datasets:

Languages:
English
Size:
n>1T
ArXiv:
License:

Attention: Work In Progress dataset

Your request will be auto approved. For any questions, please reach out to lucas@allenai.org

I understand that this dataset is a work in progress and is subject to change. I will take note of the commit ID at the time of use, and assume responsibility for any data loss or downstream inconvenience caused by this.

Log in or Sign Up to review the conditions and access this dataset content.

🚧 WARNING ‼️ THIS DATASET IS CURRENTLY WORK IN PROGRESS 🚧

Contents might change! Make a note of the commit hash you are working against.

πŸ”“ Dolma πŸ‡ Creative Commons Common Crawl πŸ•ΈοΈ

Subset of the Common Crawl corpus containing English documents with Creative Commons licenses.

Snapshot Unicode Words Documents
CC-MAIN-2013-20 3,851,018,197 5,529,294
CC-MAIN-2013-48 4,544,197,252 6,997,831
CC-MAIN-2014-10 4,429,217,941 6,682,672
CC-MAIN-2014-15 4,059,132,873 5,912,779
CC-MAIN-2014-23 5,193,195,765 8,253,690
CC-MAIN-2014-35 4,254,690,945 6,551,673
CC-MAIN-2014-41 4,289,814,449 6,558,170
CC-MAIN-2014-42 3,986,284,741 6,144,797
CC-MAIN-2014-49 3,316,075,452 4,699,472
CC-MAIN-2014-52 4,307,765,289 6,338,983
CC-MAIN-2015-06 3,675,982,679 5,181,955
CC-MAIN-2015-11 3,932,442,900 5,438,533
CC-MAIN-2015-14 3,658,107,765 4,954,273
CC-MAIN-2015-18 4,451,734,946 6,319,757
CC-MAIN-2015-22 4,285,945,319 5,949,267
CC-MAIN-2015-27 3,639,904,128 4,975,152
CC-MAIN-2016-07 1,588,496,703 3,798,207
CC-MAIN-2016-18 3,228,754,200 4,446,815
CC-MAIN-2016-22 3,217,827,676 4,242,762
CC-MAIN-2017-04 3,852,699,213 5,239,605
CC-MAIN-2017-09 4,186,915,498 5,119,171
CC-MAIN-2017-13 4,950,110,931 5,923,670
CC-MAIN-2017-17 4,684,050,830 5,645,725
CC-MAIN-2017-22 4,683,569,278 5,514,717
CC-MAIN-2017-26 4,744,689,137 5,514,047
CC-MAIN-2017-51 1,981,004,306 2,529,289
CC-MAIN-2018-13 4,816,417,930 5,520,099
CC-MAIN-2018-22 3,921,533,251 4,401,956
CC-MAIN-2018-26 4,506,583,931 4,916,546
CC-MAIN-2018-30 4,936,722,403 5,282,886
CC-MAIN-2018-34 3,865,953,978 3,808,725
CC-MAIN-2018-47 3,933,439,841 3,637,947
CC-MAIN-2018-51 4,745,124,422 4,616,832
CC-MAIN-2019-04 4,475,679,190 4,140,277
CC-MAIN-2019-09 4,287,868,800 4,142,190
CC-MAIN-2019-13 3,966,330,348 3,849,631
CC-MAIN-2019-30 4,179,526,188 4,430,572
CC-MAIN-2019-35 5,144,426,270 5,048,106
CC-MAIN-2019-39 4,572,972,457 4,527,430
CC-MAIN-2020-29 5,200,565,501 4,984,248
CC-MAIN-2020-34 4,458,827,947 4,297,009
CC-MAIN-2021-17 1,768,757,386 1,824,942
CC-MAIN-2021-39 4,599,961,675 4,287,356
CC-MAIN-2021-43 5,337,349,331 5,304,846
CC-MAIN-2021-49 3,980,018,773 4,050,641
CC-MAIN-2022-05 4,517,850,019 4,503,863
CC-MAIN-2023-06 5,135,614,227 4,959,915
CC-MAIN-2023-14 5,117,143,765 4,675,097
CC-MAIN-2023-23 5,461,486,807 4,869,627
CC-MAIN-2023-50 5,881,860,014 4,901,306
CC-MAIN-2024-10 5,164,171,562 4,335,071
CC-MAIN-2024-18 4,745,457,054 3,949,186
Total 221,715,271,483 259,728,610

Filtering Steps

Each snapshot was processed as follows:

  • Used Creative Commons extractor from the Dolma Toolkit to filter pages with Creative Commons or Public Domain license. This approach was addapted from the C4Corpus porject by Habernal et al (2016)
  • Deduplication on exact URL match
  • Filtering using subset C4 heuristic: removed pages containing javascript, lorem ipsum, and curly braces.
  • Filtering using Gopher quality and duplication heuristics.
  • Dolma repetition heuristics: removed of documents containing more than than repeated n-grams.
  • Removed any page with non-commercial (NC) or no-derivatives (ND) variants of Creative Common licenses.
  • Removal of pages with 80% duplicated ngrams.

Licensing Information

This collection is released under the Open Data Commons Attribution License (ODC-By) v1.0 license. The use of this dataset is also subject to CommonCrawl's Terms of Use.

For license extracted from individual documents in this collection, please see metadata.attribute_spans field in the dataset.

Downloads last month
673