Add CC BY-SA 3.0 Polish cultural-accessibility guides

#27
Fabryka AI org
•
edited 12 days ago

Add Polish cultural-accessibility guides

Adds kultura_bez_barier_pl: 9 records, 316 extracted pages and 217,932 measured cl100k_base proxy tokens from the official Fundacja Kultury bez Barier publication catalogue.

Source dataset: https://huggingface.co/datasets/PiotrSty/kultura-bez-barier-pl/tree/4e447551196f7f7ebebd8ceaf3fc655f42dd7e93

What the texts are

Genre: accessibility guides and handbooks for cultural institutions — practical checklists ("Warsztaty kulinarne dla osób niewidomych"), rules/how-to guides ("Napisy dla osób niesłyszących", "Przewodnik po dostępności"), a strategic report ("Model dostępnej kultury 2023"), study-visit summaries. Register: prepared didactic/institutional Polish — clear instructive prose, lists, checklists, directory-style contact sections; professionally edited, aimed at cultural-sector staff. Time span: PDF creation dates 2017-11-20 – 2024-04-19. Born-digital PDFs, not OCR'd, not machine-translated.

License at the origin — and per-guide coverage

Origin: the foundation's own publication catalogue https://kulturabezbarier.org/publikacje/. Its statement, verbatim: "Poniżej znajdują się publikacje autorstwa Fundacji Kultury bez Barier, takie jak przewodniki i poradniki pomocne w pracy z osobami z niepełnosprawnościami. Wśród nich znajdują się piktogramy ilustrujące dostępność. Oddajemy je na licencji Creative Commons (CC BY-SA 3.0). Oznacza to, że można ich używać (podając autora), zmieniać (wskazując te zmiany) i rozpowszechniać (na tej samej licencji co plik pierwotny, czyli nasz)."

Reading (stated openly, since the pronoun sits next to "piktogramy"): the grant "Oddajemy je na licencji…" with distribution terms about "plik pierwotny" covers the offered foundation-authored publication files, with the pictograms highlighted as included. This is catalogue-level evidence — no per-PDF embedded assertion is claimed (it is recorded as license_scope_note per record). The catalogue HTML is pinned in the source snapshot.

Per-guide coverage for the 9 retained records: each is a foundation-authored entry of that catalogue, with authorship corroborated per record by the catalogue statement plus PDF metadata/credit lines (attribution.jsonl). The 3 other catalogue items were rejected on rights/authorship grounds: both volumes of ABC gość niepełnosprawny name Narodowy Instytut Muzealnictwa i Ochrony Zbiorów (NIMOZ) in the PDF copyright notices (the foundation is a project partner there), and Audiodeskrypcja - zasady was dropped for legacy-font extraction damage. A third-party-authorship red flag removes the item rather than being argued away.

Sample (verbatim, from the built shard)

Practical checklist:

id: kultura_bez_barier_pl_checklisty_warsztaty_kulinarne
source: kultura_bez_barier_pl
added: 2026-09-07
created: 2024-04-19
token_count: 1438
license: CC-BY-SA-3.0
author: Julianna Matuszewska

WARSZTATY KULINARNE
DLA OSÓB NIEWIDOMYCH
#checklista

WARSZTATY KULINARNE DLA
OSÓB NIEWIDOMYCH wybierz miejsce z dużą ilością przestrzeni i dobrą akustyką zaangażuj osoby asystujące ustaw stoły / stanowiska oraz produkty według jednego schematu ustaw stanowiska w taki sposób, żeby osoba prowadząca warsztat widziała wszystkich zadbaj o dużą ilość wolnej przestrzeni na stanowiskach nie stawiaj niczego przy krawędziach stanowiska; ##MMOOŻŻEESSZ Z-:
przygotować ograniczniki np. listwy lub ustawić stoły pod ścianą zaplanuj schematyczny porządek pracy podzielony na małe kroki
/ etapy zabezpiecz noże i ostre przyrządy, a jeśli będziesz korzystać ze źródeł ciepła (piec, woda), spytaj, czy ktoś potrzebuje pomocy opisz osobom biorącym udział przestrzeń warsztatów oraz stanowiska - układ itp.
opisz […]

Rules/how-to guide (the doubled letters in the checklist above are visible PDF layout residue; this one is clean prose):

id: kultura_bez_barier_pl_napisy_dla_nieslyszacych
source: kultura_bez_barier_pl
added: 2026-09-07
created: 2019-01-31
token_count: 7687
license: CC-BY-SA-3.0
author: Izabela Künstler; Urszula Butkiewicz

NAPISY
DLA OSÓB NIESŁYSZĄCYCH I SŁABOSŁYSZĄCYCH – zasady tworzenia
WSTĘP
Napisy dla niesłyszących i słabosłyszących tworzy się, aby umożliwić osobom z dysfunkcją słuchu jak najpełniejszy odbiór utworów audiowizualnych – zapewnić tym grupom widzów nie tylko zrozumienie treści filmów, spektakli teatralnych, programów telewizyjnych itp., ale też satysfakcję z ich odbioru.
Cele te można osiągnąć poprzez jak najdoskonalsze odwzorowanie ścieżki dźwiękowej w słowie. Dlatego oprócz dialogów napisy powinny zawierać: niewerbalne wypowiedzi bohaterów (okrzyki i pomruki wyrażające emocje), informacje o istotnych dźwiękach, odgłosach i tle muzycznym oraz informacje ułatwiające identyfikację bohaterów. Poza specyficznymi zasadami, wynikającymi z potrzeb odbiorców, w napisach dla osób niesłyszących i […]

Strategic report (multi-institutional, expert lists):

id: kultura_bez_barier_pl_model_dostepnej_kultury_2023
source: kultura_bez_barier_pl
added: 2026-09-07
created: 2023-12-20
token_count: 73667
license: CC-BY-SA-3.0
author: Fundacja Kultury bez Barier; Robert Więckowski

Model dostępnej kultury

Model dostępnej kultury
Warszawa, 2023 rok
Eksperci biorący udział w projekcie:
(alfabetycznie)
Fundacja Kultury bez Barier
Michalina Adamska
Marta Matowicka
Emilia Piegat
Marta Piwowarska
Alicja Szurkiewicz
Robert Więckowski
Dariusz Zajkowski
Anna Żórawska
Ministerstwo Kultury i Dziedzictwa Narodowego
Radosław Dubisz
Małgorzata Dzik
Agnieszka Kozik
Magdalena Paluch – Wojtas
Artur Witkowski
Państwowy Fundusz Rehabilitacji Osób Niepełnosprawnych
Anna Bogusz – Koźbiał
Natalia Pamuła
Alina Szklaruk
Irena Wolińska
Centrum Kompetencji w Zakresie Kreatywnego Starzenia się i Sztuki Włączającej
(kubia)
Annette Ziegert
Annalena Knors
Lisette Reuter
Zespół Konsultacyjny:
Magdalena Karowska-Koperwas
Justyna Mańkowska
Tomasz Włodarski
Korekta: Emilia Piegat
Oprawa graficzna i […]

PII — which scrubber ran

Pattern redaction in src/build_kultura_bez_barier_pl.py (e-mail -> [REDACTED:EMAIL], labelled phone -> [REDACTED:PHONE], limited patterns) — not SlayerLab/NERGAL. Spot check over all 9 records: 14 e-mail redactions confirm the scrubber works, but real residual contacts survive in kultura_bez_barier_pl_podsumowanie_wizyt_studyjnych — a published directory of museum accessibility coordinators where names and phones stay while the e-mail beside them is redacted: "Emila Kamola Tel./SMS + 48 509 490 034 e-mail: [REDACTED:EMAIL]", "Katarzyna Lubas Tel./SMS: 535 36 37 09", "Tel/SMS: 504449027". The label-anchored phone pattern misses Tel./SMS / Tel/SMS variants and unlabelled numbers; names are out of its scope entirely. Other hits benign: 8 museum addresses, 17 instructive mentions of "telefon", one Facebook URL with an id. Published professional contacts in a CC BY-SA guide — but identifiable individuals' data. No release before a NERGAL pass (it must catch what patterns structurally cannot); merge with release: None is fine meanwhile.

Rights and method (original)

The official catalogue describes these linked foundation-authored publications as CC BY-SA 3.0 (statement above). The contribution preserves the catalogue HTML, raw PDFs, URLs, hashes, attribution and modification notice. This is catalogue-level evidence; it does not assert that every PDF embeds a separate license notice. Text was extracted with pdfplumber, normalized, checked for limited PII patterns, and exact/near-deduplicated within source. Selection: 12 discovered → 9 retained (2 rejected on rights/authorship, 1 on extraction quality).

Slayer ontology and remaining gates

The manifest separates content-addressed source/dataset Versions, processing and validation Protocols, executed Runs, Evidence, falsifiable Claims, Actors and typed lineage. Remaining release gates: NERGAL pass, maintainer sign-off on the catalogue-level scope reading, cross-source DynaWord deduplication, benchmark-overlap testing and controlled training ablations. PDF extraction may flatten tables and omit visual meaning. This PR proposes a source, not a merged or stable release.

Registry and test plan

  • Registry entry unchanged — kultura_bez_barier_pl is already in src/sources.py on main; this PR now contains only the source files (the earlier merge-based branch had stale-branch drift — it is replaced by a single clean commit on main @ 34e7ea3)
  • python3 -m pytest src/test_kultura_bez_barier_contribution.py — 2 passed
  • Residual-PII spot check over all 9 records (above, real residual contacts reported)
  • Reviewer: catalogue-level license reading accepted
  • Reviewer: NERGAL-before-release note accepted
Fabryka AI org

Cannot merge
This branch has merge conflicts in the following files:
src/sources.py

Fabryka AI org

@kacperwikiel Konflikt w src/sources.py rozwiązany — scaliłem aktualny main do gałęzi PR (commit a7d8716). Wpis kultura_bez_barier_pl jest już na main, więc PR nie zmienia sources.py, a jedynie dodaje pliki źródła (data/kultura_bez_barier_pl/*, src/build_kultura_bez_barier_pl.py, test kontraktowy). Testy kontraktowe przechodzą. Gotowe do merge — dzięki!

Fabryka AI org

Review against AGENTS.md ("What the PR must show", "Normalize the text before you build"). Needed before merge:

  1. Registry entry. kultura_bez_barier_pl is already in src/sources.py on main — drop it from this PR.
  2. License at the origin. Link the license statement on the catalogue's own page, not only your HF mirror — the mirror can't be evidence for itself. Confirm the catalogue-level license covers each of the 9 guides.
  3. Describe the texts — genre, register, time span.
  4. Sample inline. Paste 2–3 real documents into the description, one fenced block each: metadata as key: value lines, a blank line, then the text with its real line breaks, truncated with […] (format as in #24). A linked .sample.jsonl renders on the Hub as one escaped line per document and can't be reviewed at a glance.
  5. PII. AGENTS.md requires SlayerLab/NERGAL. The source can merge with release: None after another scrubber, but it won't be admitted to a release until the shard is re-run through NERGAL. State in the datasheet which scrubber ran.
Fabryka AI org

Review addressed — the branch is replaced by a single clean commit on main @ 34e7ea3. Point by point:

  1. Registry entry dropped. kultura_bez_barier_pl stays in src/sources.py on main untouched; this PR now changes only the source's own files (data/kultura_bez_barier_pl/*, src/build_kultura_bez_barier_pl.py, src/test_kultura_bez_barier_contribution.py, artifacts/kultura_bez_barier_pl_ontology_manifest.json, artifacts/source_findings.md entry). Note for the record: the previous tip was a merge of an old main whose tree resurrected scrub_pii.py/scrub_entities.py and deleted render_sample.py — the clean commit drops all of that drift, so "Ready to merge" is now actually true.
  2. License at the origin. The description and datasheet now quote the catalogue's own statement at https://kulturabezbarier.org/publikacje/ verbatim and state the reading openly (the "Oddajemy je na licencji…" grant with the "plik pierwotny" distribution terms covers the offered foundation-authored publication files; the pronoun sits next to "piktogramy", so this is catalogue-level, not per-PDF — exactly what license_scope_note records). Per-guide coverage confirmed for all 9: each is a foundation-authored catalogue entry with authorship corroborated by PDF metadata/credit lines in attribution.jsonl, and the 3 other catalogue items were rejected on rights/authorship grounds (2 NIMOZ co-published volumes whose PDF copyright notices name the third-party institute; 1 for legacy-font extraction damage) — a third-party authorship red flag removes the item rather than being argued away.
  3. Texts described. Genre (accessibility checklists, rules/how-to guides, a strategic report, study-visit summaries), register (prepared didactic/institutional Polish for cultural-sector staff), time span (PDF creation dates 2017-11-20 – 2024-04-19, per-record created_semantics recorded).
  4. Sample inline. 3 real documents in the description, one fenced block each (key: value metadata, blank line, real line breaks, […] truncation, format as in #24): a checklist (with visible PDF layout residue — honest), a clean rules guide, the strategic report.
  5. PII. Datasheet names the scrubber (pattern pass in src/build_kultura_bez_barier_pl.py — not NERGAL) and reports a spot check over all 9 records: 14 e-mail redactions work, but real residual contacts survive in kultura_bez_barier_pl_podsumowanie_wizyt_studyjnych — coordinator names + phones under Tel./SMS / Tel/SMS label variants the phone pattern misses, with the e-mail beside them redacted ("Emila Kamola Tel./SMS + 48 509 490 034 e-mail: [REDACTED:EMAIL]", "Katarzyna Lubas Tel./SMS: 535 36 37 09", "Tel/SMS: 504449027"). Agreed: no release before a NERGAL pass — this is exactly the case the model-based scrubber exists for; merge with release: None is fine meanwhile.

Verification: python3 -m pytest src/test_kultura_bez_barier_contribution.py — 2 passed.

ppuzio changed pull request status to merged
ppuzio deleted the refs/pr/27 ref

Sign up or log in to comment