LICENSES.md

Licences

• ja_aozorabunko
  • Texts from Aozora Bunko
  • Primary distribution: https://huggingface.co/datasets/globis-university/aozorabunko-clean
  • License: CC BY 4.0
• ja_cc
  • A Japanese corpus that was extracted and filtered from the whole of Common Crawl
  • The filtering process was conducted using Uzushio.
    • level0: No duplicate pages
    • level1: May contain a small amount of duplicate content, but mostly unique pages
    • level2: Another set, similar to level1, may contain some duplicates, but mostly unique
  • The filtering settings used can be found here.
  • For more details, refer to these Qiita articles.
  • License: ODC-BY (Users must also comply with Common Crawl terms of use.)
• ja_e-gov
  • e-Gov legal texts
  • Primary distribution: https://huggingface.co/datasets/nlp-waseda/e_gov
  • License: CC BY 4.0
• ja_fineweb-2
  • Japanese portion of FineWeb 2, a multilingual dataset filtered from Common Crawl
  • Primary distribution: https://huggingface.co/datasets/HuggingFaceFW/fineweb-2
  • License: ODC-BY (Users must also comply with Common Crawl terms of use.)
• ja_kaken
  • Text of outline for each research project from KAKEN (Grants-in-Aid for Scientific Research Database)
  • License: CC BY 4.0
• ja_kokkai_giji
  • Japanese Diet minutes texts obtained and extracted using the API of Full-text Database System for the Minutes of the Diet
  • License: CC BY 4.0
• ja_nwc2010
  • A Japanese corpus extracted and filtered from the HTML archive of Japanese Web Corpus 2010 using Uzushio
  • License: CC BY 4.0
• ja_patent
  • Patent texts extracted from the Japanese Patent Office's published data files
  • License: CC BY 4.0
• ja_sip_comprehensive_html/ja_sip_comprehensive_pdf
  • The general web corpus constructed in the 3rd phase of the Strategic Innovation Promotion Program (SIP), "Utilization of Generative AI in Building Integrated Healthcare Systems"
  • Primary distribution: https://gitlab.med-jp.nii.ac.jp/datasets/sip3-ja-general-web-corpus
  • License: CC BY 4.0
• ja_warp_html/ja_warp_pdf
  • Texts crawled and extracted from websites based on URLs collected by the Web Archiving Project (WARP) of the National Diet Library (NDL). The URL list was provided by NDL.
    • ja_warp_html: Texts extracted from HTMLs
    • ja_warp_pdf/e0: Texts extracted from PDFs created with Word, TeX, or Ichitaro
    • ja_warp_pdf/e0.2: Texts extracted from PDFs created with PowerPoint, Excel, etc.
  • License: CC BY 4.0
• ja_wiki
  • Japanese Wikipedia (20230720 version)
  • License: CC BY-SA 3.0