# govpl

gov.pl — Polish government press releases

## Dataset description
- **Source (upstream):** https://www.gov.pl/
- **Domain:** government
- **Language:** Polish (pl)
- **License:** `CC-BY-SA-4.0`
- **Created (range):** 2005-09-26, 2026-07-19
- **Added:** 2026-07-19; PAP reprints removed 2026-10-06; other third-party texts removed 2026-10-19

## Licensing — traceable basis
gov.pl offers its text under CC BY-SA 4.0. The site footer and https://www.gov.pl/web/gov/prawa-autorskie say: "Treści tekstowe publikowane w serwisie (z wyłączeniem treści audiowizualnych), są udostępniane na licencji typu Creative Commons: uznanie autorstwa - na tych samych warunkach 4.0 (CC BY-SA 4.0)." All rows rely on this offer as it stood when they were fetched (2026-07-19). Wayback captures of the page from 2026-07-18 (https://web.archive.org/web/20260718065324/https://www.gov.pl/web/gov/prawa-autorskie) and 2026-10-15 (https://web.archive.org/web/20261015193442/https://www.gov.pl/web/gov/prawa-autorskie) show the same text. Earlier footers were different: CC BY 3.0 PL from the first capture (2019-08-05) to 2022-08-17, CC BY-NC-ND 3.0 PL from 2022-08-19 to 2025-12-12, and today's text from 2025-12-18. By the date listed on gov.pl, 40,708 rows date from the CC BY-NC-ND 3.0 PL period, 26,974 from the CC BY 3.0 PL period and 9,910 from before 2019-08-05. That today's offer also covers texts first published under an earlier footer is our reading; it is not certain. The same page excludes works whose copyright belongs to third parties; using those needs the owners' consent. Articles crediting PAP and other third-party texts are therefore dropped (legal note).

## Provenance
Ministry/agency press releases fetched directly from gov.pl (editor-content per article, HTML stripped) by src/fetch_govpl.py. Subsites + their non-uniform news-section paths are enumerated by src/discover_govpl.py into the committed manifest src/govpl_subsites.json (gov.pl has no sitemap/API). v0.2.6 removes PAP reprints and the third-party texts listed in govpl.third-party-review.csv from the v0.2.5 Parquet with src/clean_govpl.py, which also writes govpl.attribution.jsonl: each row's gov.pl URL, subsite, title and date.

## Attribution
CC BY-SA 4.0 requires credit and a link to the source. `govpl.attribution.jsonl` has one
line per row, in Parquet order: `id`, `url` (the gov.pl article), `subsite` (the publishing
ministry or agency), `title`, `date` (as listed on gov.pl) and `license`. Rows were matched
to the 2026-07-19 crawl on SHA-1 of the stripped text; for 1,185 rows the same text was
fetched from more than one URL, and the first in fetch order is given.

## Statistics
| documents | characters | tokens (tiktoken proxy) |
|---:|---:|---:|
| 87,816 | 216,869,817 | 80,204,973 |

## Per-document license metadata
| license | documents |
|---|---:|
| `CC-BY-SA-4.0` | 87,816 |

Author metadata present for **0** documents. Empty values mean the upstream record did not expose a machine-readable author field.

## Legal scope note
gov.pl's copyright page carves out "utwory, co do których prawa autorskie przysługują osobom trzecim" from its CC BY-SA 4.0 grant. Articles crediting PAP (Polish Press Agency) as their source are such works, so 188 are dropped (rule fetch_govpl.pap_rule, applied by src/clean_govpl.py): a "Źródło:" line naming PAP (163), "Na podst. … PAP" (22), a "(PAP)" sign-off line (2) and one PAP interview transcript (1). PAP photo credits and articles that only quote or mention PAP stay. Other third-party text was reviewed in v0.2.6 (our assessment, src/clean_govpl.py with govpl.third-party-review.csv): every other article with a "Źródło:" line (710) was classified by the source the line credits, and interview and reprint rules found 6 more. 186 are dropped: 13 whose text is credited, wholly or partly, to a news portal, news agency, broadcaster or magazine, 167 whose text is credited, wholly or partly, to a source other than a public office (for example an NGO, an international or EU body, a company, a university, a museum or another website) and 6 interviews and reprints from other outlets (one names no outlet). The 530 that stay credit their text to a public office (436) or credit only data, a chart, a reference or a quotation (94). Teasers that quote a third-party interview and point to it stay. Articles that match none of these rules were checked on a seeded sample of 200; none was a third-party text.

## Filters applied
`build_dynaword.py` (v0.2.2), minimal per Dynaword guidelines:
- drop documents < 200 chars: **0**
- drop non-Polish (diacritic ratio): **24**
- exact cross-source dedup (sha1): **1,926**
- read 90,140 → kept 88,190

`src/clean_govpl.py` (v0.2.6), third-party carve-out:
- drop PAP reprints: **188** (`Źródło:` line 163, `Na podst.` 22, `(PAP)` sign-off 2, interview 1)
- drop other third-party texts: **186** (`govpl.third-party-review.csv`: text credited to media 13, to other non-public sources 167, interviews and reprints 6)
- read 88,190 → kept 87,816

Token counts are a fast tiktoken (cl100k) proxy (~1% off Llama-3); the canonical
Llama-3 count is computed at release.