Address review: origin license evidence, BY-SA treatment, datasheet provenance

- NOTICE/datasheet: cite the CC BY-SA 3.0 statement at the origin (per-lesson
Licencja line + site footer + XML rights fields), not the mirror.
- Datasheet: texts described (genre, register, time span 2012-11-09 - 2015-01-12);
collection/normalization reported with gates and drops; inclusion argument
and caveats.
- Corpus-level CC BY-SA 3.0 treatment proposed: per-record attribution fields
and a share-alike note (adaptation released under CC BY-SA 4.0).
- artifacts/source_findings.md: edukacja_medialna_pl finding appended.

artifacts/source_findings.md

@@ -611,3 +611,40 @@ the card: single term (10th), private citizens named.
 belongs in meta, not text: stripping it makes the body comparable across
 question and reply docs, and the before/after is the fastest reviewer evidence
 for any Sejm HTML source.
+
+---
+
+## Edukacja Medialna — media-literacy lessons (`edukacja_medialna_pl`)
+
+**Investigated** — 2026-09-06 (review round 2026-09-28).
+
+**Verdict** — `added`. 252 records / 728,987 `cl100k_base` tokens of CC BY-SA
+3.0 lesson materials — small, but a school-teaching register nothing else in the
+corpus covers.
+
+**Enumeration** — <https://edukacjamedialna.edu.pl/lekcje/> lists every lesson
+(255 at fetch). Each lesson page carries its source XML: a `utwor` document with
+Dublin Core fields and the authored `powiesc` body. `src/build_edukacja_medialna_pl.py`
+fetches page + XML, checksums both (`page_sha256`, `xml_sha256`) and builds from
+the pinned `artifacts/raw_xml.jsonl.gz`; nothing else is needed.
+
+**License evidence** — verified at the origin (the mirror is not evidence for
+itself): every lesson page ends with `Licencja: Creative Commons Uznanie
+autorstwa - Na tych samych warunkach 3.0` linking
+<https://creativecommons.org/licenses/by-sa/3.0/> (e.g.
+<https://edukacjamedialna.edu.pl/lekcje/abc-bezpieczenstwa-w-sieci/>), the site
+footer repeats the grant (<http://creativecommons.org/licenses/by-sa/3.0/deed.pl>),
+and the source XML carries `dc:rights` + `dc:rights.license`. The build gate
+requires the statement on page AND XML. Per-record `rights`/`rights_urls` and
+hashes are pinned in `attribution.jsonl`.
+
+**Blockers** — none to merge with `release: None`. Before a release: NERGAL pass
+(the in-build redaction is e-mail/labelled-phone patterns only — 1 match), cross-source
+exact/near dedup, benchmark-overlap check, and maintainer review of embedded
+third-party excerpts and of the corpus-level BY-SA 3.0 treatment (proposal in the
+datasheet: per-record attribution + share-alike note, adaptation released under
+CC BY-SA 4.0).
+
+**Lesson** — cite the license where the publisher states it (per-lesson
+`Licencja:` line + XML rights fields), never a mirror. Pin the statement with
+`page_sha256`/`rights_urls` so it survives site redesigns.

data/edukacja_medialna_pl/NOTICE.md

@@ -3,6 +3,7 @@
 Source: https://edukacjamedialna.edu.pl/lekcje/
 Publisher: Fundacja Nowoczesna Polska.
 License: CC BY-SA 3.0 (https://creativecommons.org/licenses/by-sa/3.0/).
+License statement at the origin: every lesson page ends with "Licencja: Creative Commons Uznanie autorstwa - Na tych samych warunkach 3.0" (e.g. https://edukacjamedialna.edu.pl/lekcje/abc-bezpieczenstwa-w-sieci/), and the edukacjamedialna.edu.pl site footer repeats the grant ("Jeśli nie oznaczono inaczej, wszystkie materiały na stronie są objęte wolną licencją ... – Na tych samych warunkach 3.0").
 
 Per-record text/scenario/expert creators, canonical URLs, source dates and rights evidence are preserved in `artifacts/attribution.jsonl`. Raw source XML and acquisition hashes are preserved in `artifacts/raw_xml.jsonl.gz`.
 

data/edukacja_medialna_pl/edukacja_medialna_pl.md

@@ -23,6 +23,97 @@ A text-only snapshot of Polish media-literacy lesson explanations and scenarios
 
 See `NOTICE.md`, `artifacts/attribution.jsonl`, `artifacts/decisions.jsonl`, `artifacts/qa.json`, `artifacts/checksums.json` and `artifacts/ontology.json`. Twelve complete deterministic records are in `artifacts/sample.jsonl`.
 
+## Origin and license
+
+- Origin: Edukacja Medialna, run by Fundacja Nowoczesna Polska — lesson index
+  <https://edukacjamedialna.edu.pl/lekcje/>, per-lesson source XML pinned in
+  `artifacts/raw_xml.jsonl.gz`.
+- License statement **at the origin** (verified on the site, not on the mirror):
+  every lesson page ends with `Licencja: Creative Commons Uznanie autorstwa -
+  Na tych samych warunkach 3.0`, linking <https://creativecommons.org/licenses/by-sa/3.0/>
+  — e.g. <https://edukacjamedialna.edu.pl/lekcje/abc-bezpieczenstwa-w-sieci/> —
+  and the site footer repeats the grant: "Jeśli nie oznaczono inaczej, wszystkie
+  materiały na stronie są objęte wolną licencją Creative Commons Uznanie
+  autorstwa – Na tych samych warunkach 3.0"
+  (<http://creativecommons.org/licenses/by-sa/3.0/deed.pl>).
+- The build gate requires explicit CC BY-SA 3.0 on both the lesson page and the
+  source XML (`dc:rights` + `dc:rights.license`); per-record evidence (`rights`,
+  `rights_urls`, `page_sha256`, `xml_sha256`) is in `artifacts/attribution.jsonl`.
+
+## What the texts are
+
+- **Genre**: authored teaching materials — knowledge-explainer prose ("Wiedza w
+  pigułce"), lesson plans and projects ("Pomysł na lekcję", "Zadanie": objectives,
+  classroom procedure, evaluation), and glossary entries ("Słowniczek"). One
+  document per lesson.
+- **Register**: prepared didactic Polish — formal written prose with classroom
+  instructions, exercises and check questions; gender-inclusive paired forms
+  ("uczestnicy i uczestniczki"); standard spelling. Human-authored, not OCR'd,
+  not machine-translated; 62 credited authors across 252 records (roles
+  `textbook`, `scenario`, `expert`, `methodologist`).
+- **Time span**: upstream publication dates 2012-11-09 – 2015-01-12 (per-record
+  `created`). The material includes historically dated technology and legal
+  guidance (see Limitations).
+
+## Collection and normalization
+
+Every step lives in `src/build_edukacja_medialna_pl.py` and runs against the
+pinned raw XML, so the shard is reproducible:
+
+- **Acquire**: lesson index → 255 lesson URLs; lesson page HTML and source XML
+  fetched and checksummed (`page_sha256`, `xml_sha256`).
+- **Extract**: only the authored `powiesc` body; leaf blocks kept (headings,
+  paragraphs, list points, glossary definitions, activity/scenario fields),
+  nested container duplication skipped. Extraction stops at the "Czytelnia"
+  heading — external reading lists, bibliography and long-quote elements are
+  omitted.
+- **Normalize**: NFKC; soft hyphens (U+00AD), zero-width spaces (U+200B) and
+  control characters removed; space/tab/NBSP runs collapsed per line, lines
+  trimmed; runs of 3+ newlines capped at one empty line (paragraph breaks kept).
+- **PII**: pattern-based redaction only — e-mail addresses → `[REDACTED:EMAIL]`,
+  labelled phone numbers (`telefon`/`tel.`) → `[REDACTED:PHONE]`. Matches in the
+  shard: 1 e-mail, 0 phones. This is not de-identification; a NERGAL pass is
+  required before release.
+- **Gates and drops**: explicit CC BY-SA 3.0 on page and XML (255/255), credited
+  creator (3 dropped: `missing_creator`), body ≥ 300 chars (0 dropped), langid
+  Polish (7-language set, confidence ≥ 0.99; 0 dropped), exact dedup on
+  normalized text (0 dropped), near-dedup seeded MinHash 128 / LSH 0.8 with
+  exact 5-word Jaccard ≥ 0.9 (0 dropped). 252 of 255 retained.
+
+Before/after excerpts of the extraction and normalization are in the PR
+description.
+
+## Corpus-level CC BY-SA 3.0 treatment (proposal)
+
+The records are CC BY-SA 3.0 works; the corpus is redistributed as CC BY-SA 4.0.
+
+- **Attribution, per record**: `author` in the parquet plus
+  `artifacts/attribution.jsonl` — `authors` with `creator_roles`, `title`,
+  `canonical_url`/`page_url`/`xml_url`, `publisher`, `created`, `license` +
+  `license_url`, source `rights`/`rights_urls`, and
+  `page_sha256`/`xml_sha256`/`text_sha256`. Enough to credit each document under
+  BY-SA 3.0 §4 without consulting the mirror.
+- **Share-alike note**: the corpus is an adaptation and is released under
+  CC BY-SA 4.0 — a later version of the same License Elements, which
+  [BY-SA 3.0 §4(b)](https://creativecommons.org/licenses/by-sa/3.0/legalcode)
+  permits for adaptations. Redistributors of these records or adaptations of
+  them must keep the per-record attribution and pass share-alike terms on
+  (CC BY-SA 3.0 or a later version with the same License Elements).
+
+## Inclusion argument
+
+What this adds: authored didactic prose — objectives, exercises, classroom
+procedure, glossary definitions — a school-teaching register the corpus does not
+otherwise draw from (its registers are encyclopedic, legal, parliamentary,
+journalistic and fiction). 728,987 tokens is a diversity addition, not a scale
+one.
+
+Bias and quality caveats: a single publisher (Fundacja Nowoczesna Polska) and 62
+credited authors; a school-audience framing with gender-inclusive paired forms;
+technology and legal guidance frozen at 2012–2015; embedded third-party excerpts
+(quotations, linked works) may need separate review before release. Training
+benefit is an untested diversity hypothesis.
+
 ## Limitations
 
 The source includes historically dated technology and legal guidance. Dates are upstream publication metadata, not proof of present-day currency. External reading lists are omitted. Embedded excerpts may require additional review. PII checks are pattern-based, not comprehensive de-identification. Cross-source DynaWord deduplication and benchmark-overlap checks remain pending. Training benefit is an untested hypothesis requiring controlled ablations.