README.md

metadata

language:
  - en
license: odc-by
tags:
  - document-classification
  - text-classification
  - web-pages
configs:
  - config_name: AcademicWriting
    data_files:
      - path:
          - AcademicWriting.jsonl.zst
        split: train
  - config_name: All
    default: true
    data_files:
      - path:
          - All/train.jsonl.zst
        split: train
      - path:
          - All/test.jsonl.zst
        split: test
  - config_name: ContentListing
    data_files:
      - path:
          - ContentListing.jsonl.zst
        split: train
  - config_name: CreativeWriting
    data_files:
      - path:
          - CreativeWriting.jsonl.zst
        split: train
  - config_name: CustomerSupportPage
    data_files:
      - path:
          - CustomerSupportPage.jsonl.zst
        split: train
  - config_name: DiscussionForumCommentSection
    data_files:
      - path:
          - DiscussionForumCommentSection.jsonl.zst
        split: train
  - config_name: FAQs
    data_files:
      - path:
          - FAQs.jsonl.zst
        split: train
  - config_name: IncompleteContent
    data_files:
      - path:
          - IncompleteContent.jsonl.zst
        split: train
  - config_name: KnowledgeArticle
    data_files:
      - path:
          - KnowledgeArticle.jsonl.zst
        split: train
  - config_name: LegalNotices
    data_files:
      - path:
          - LegalNotices.jsonl.zst
        split: train
  - config_name: Listicle
    data_files:
      - path:
          - Listicle.jsonl.zst
        split: train
  - config_name: NewsArticle
    data_files:
      - path:
          - NewsArticle.jsonl.zst
        split: train
  - config_name: NonfictionWriting
    data_files:
      - path:
          - NonfictionWriting.jsonl.zst
        split: train
  - config_name: OrganizationalAboutPage
    data_files:
      - path:
          - OrganizationalAboutPage.jsonl.zst
        split: train
  - config_name: OrganizationalAnnouncement
    data_files:
      - path:
          - OrganizationalAnnouncement.jsonl.zst
        split: train
  - config_name: PersonalAboutPage
    data_files:
      - path:
          - PersonalAboutPage.jsonl.zst
        split: train
  - config_name: PersonalBlog
    data_files:
      - path:
          - PersonalBlog.jsonl.zst
        split: train
  - config_name: ProductPage
    data_files:
      - path:
          - ProductPage.jsonl.zst
        split: train
  - config_name: QAForum
    data_files:
      - path:
          - QAForum.jsonl.zst
        split: train
  - config_name: SpamAds
    data_files:
      - path:
          - SpamAds.jsonl.zst
        split: train
  - config_name: StructuredData
    data_files:
      - path:
          - StructuredData.jsonl.zst
        split: train
  - config_name: TechnicalWriting
    data_files:
      - path:
          - TechnicalWriting.jsonl.zst
        split: train
  - config_name: TranscriptInterview
    data_files:
      - path:
          - TranscriptInterview.jsonl.zst
        split: train
  - config_name: TutorialHow-ToGuide
    data_files:
      - path:
          - TutorialHow-ToGuide.jsonl.zst
        split: train
  - config_name: UserReviews
    data_files:
      - path:
          - UserReviews.jsonl.zst
        split: train

English Document Format Classification Dataset

English-language web pages classified by document type, designed to train robust text classifiers and provide ready-to-use data for specific web formats.

• Purpose: Train generalized document classifiers or extract clean, single-format corpora for specific downstream tasks.
• Configurations: Each document type is available in its own dedicated dataset configuration (e.g., TutorialHow-ToGuide, PersonalAboutPage).
• Splits: The All configuration contains every document type combined, featuring a 10% stratified test split.

Curation & Filtering Pipeline

This dataset is derived from a subset of the first 1 million rows of allenai/c4, utilizing annotations from agentlans/en-document-classification.

1. Filtering: Samples are included only when the weborganizer_format field matches doc_type_v2_primary.
2. Personally identifiable information (PII) removal: Samples were removed based on agentlans/multilingual-e5-small-pii-detector.

Class Distribution (All Split)

Label	Train	Test	Total
Academic Writing	7,913	879	8,792
Content Listing	10,204	1,134	11,338
Creative Writing	2,947	328	3,275
Customer Support Page	5,237	582	5,819
Discussion Forum / Comment Section	3,802	423	4,225
FAQs	1,505	167	1,672
Incomplete Content	758	84	842
Knowledge Article	17,141	1,905	19,046
Legal Notices	3,307	368	3,675
Listicle	4,546	505	5,051
News Article	35,401	3,933	39,334
Nonfiction Writing	830	92	922
Organizational About Page	17,953	1,995	19,948
Organizational Announcement	21,876	2,431	24,307
Personal About Page	6,514	724	7,238
Personal Blog	40,259	4,473	44,732
Product Page	146,827	16,314	163,141
Q&A Forum	4,069	452	4,521
Spam / Ads	67	7	74
Structured Data	2,110	235	2,345
Technical Writing	3,643	405	4,048
Transcript / Interview	777	86	863
Tutorial / How-To Guide	20,244	2,249	22,493
User Reviews	6,563	729	7,292
Total	364,493	40,500	404,993

Licensing

Distributed under the Open Data Commons Attribution License (ODC-BY), matching the licensing terms of upstream sources allenai/c4 and agentlans/en-document-classification.