Datasets:
README: copied Wikipedia text and the Trędowata overlap (v2#80, for the v0.2.6 release commit)
#115
by ppuzio - opened
Draft, for the v0.2.6 release commit (v2#80). Adds the two card notes approved on 2026-10-06. No data changes.
- Disclaimer & legal, new bullet before Notice-and-takedown: the Wikipedia build drops pages with plwiki's copyright-violation notice but cannot find an unmarked paste. The hand-checked sample found 2 of 80 pages copied (about 2.5%; 95% Wilson upper bound 12.9% per licence label). Both are dropped; others are removed on notice. Points to
data/wikipedia/wikipedia.md, which has the method and sample. - Method, new paragraph: cross-source deduplication covers some pairs only. Below the 0.9 shingle threshold both copies stay, so both tomes of Mniszek's Trędowata (0.7588, 0.6473) are in
wolne_lekturyandeltec_pol. Points towolne_lektury.eltec_pairs.tsv.
The text is in src/make_docs.py and, by hand, in README.md (CRLF kept). Running make_docs.py on main also rewrites stale counts in README, CHANGELOG and one datasheet, which the release commit recomputes, so it is not run here. Its output contains both notes verbatim.
The takedown wording itself is lawyer question H1 (v2#56); this PR only points to the existing procedure.
Checks
- Full
pytest src(LFS not smudged): the same 119 failures/errors as ataf8ddc7, all tests that read parquet data.
🤖 Generated with Claude Code
ppuzio changed pull request status to open
ppuzio changed pull request status to merged
ppuzio deleted the
refs/pr/115 ref