README.md

metadata

license: cc-by-sa-3.0
language:
  - pl
task_categories:
  - text-generation
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-00000-of-00001.parquet

Nonsensopedia - Polish satirical wiki (content namespaces)

Content pages from the pinned wikiteam3 dump of nonsa.pl (https://archive.org/details/wiki-nonsa.pl-20251009, generated 20251009). Nonsensopedia is the Polish-language satirical encyclopedia - a community wiki parodying Wikipedia and its sister projects. Content namespaces included: 0 (articles), 100 (Cytaty - quotes), 102 (NonNews - fake news), 104 (NonZrodla - fake sources), 106 (Slownik - fake dictionary), 108 (Gra - games), 110 (Forum), 114 (Poradnik - guides). Latest revision per page. This is informal, humorous creative Polish - satire, jokes, parody reference text.

• Dump: nonsa.pl-20251009-history.xml.zst, sha1-verified against archive.org item metadata
• Non-redirect content pages: 26,781
• Retained after text QA and within-source deduplication: 12,391
• Characters: 44,870,619
• Tokens: 17,107,478 (cl100k_base proxy)
• License: CC BY-SA 3.0 per record; authorship via per-page history link

Provenance and rights

Nonsensopedia text is licensed CC BY-SA 3.0; the archive.org item metadata declares the same licenseurl. Each record keeps its page id, title, namespace, last-revision timestamp and canonical URL - the page revision history is the attribution trail. The dump was produced by the WikiTeam preservation project (wikiteam3). Additional third-party attribution notices must be preserved and remain a review item. The created field is the last revision date, not page creation.

Processing and limitations

Wikitext is stripped with mwparserfromhell; lines occurring in at least 1% of documents are removed as cross-document boilerplate (see artifacts/boilerplate_lines.json). Unicode and whitespace normalization, email/labelled-phone/IP/labelled-national-ID/account-candidate pattern redaction, three-window langid vote, exact and near deduplication within source. The final text must contain at least 500 characters after redaction. Rejection counts:

{
  "too_little_text": 14338,
  "low_letter_ratio": 23,
  "non_polish_text": 28,
  "too_little_text_after_redaction": 1
}

PII filtering masks parsed IPv6 addresses (including compressed forms), labelled PESEL/NIP/REGON and 26-digit account candidates as well as the existing patterns. The account filter is conservative and may mask non-account numbers; unlabelled identifiers, free-form phone numbers, names and personal disclosures may remain. This is not complete anonymization.

The register is deliberately non-factual: satirical articles, fake news, fake dictionary entries, quotes, forum posts and game pages. Pages may parody or quote real Wikipedia articles and other copyrighted works; quoted passages are not guaranteed disjoint from other corpus shards. The record model is a whole content page; individual entries are not split.

Review artifacts

See artifacts/sample.jsonl, attribution.jsonl, decisions.jsonl, source_manifest.jsonl, overlap_audit.json, stats.json, qa.json, checksums.json, run.json and ontology.json.