Releases: gbrlpzz/german-pareto-deck
Releases · gbrlpzz/german-pareto-deck
Release list
v0.3 - verified no-audio deck
v0.3 — verified no-audio release
- 3,122 cards: 2,560 word cards, 500 pattern cloze cards, and 62 recognition cards.
- Core: 1,313 cards. Extension: 1,247 cards. Patterns: 562 cards.
- 882 validated Kaikki lemma overrides; 6 rejected source links recorded.
- 28 gloss-less corpus/name rows removed and recorded.
- Deterministic translation ties and byte-identical APKG builds.
- No audio and no raw corpora.
See RELEASE_MANIFEST.json in the Zenodo bundle for hashes, source checksums, and test commands.
v0.2 - card quality
v0.2
- 3,525 cards: 2,963 word cards, 500 pattern cloze cards, and 62 recognition cards.
- All 500 selected patterns now have cloze cards.
- Added punctuation- and apostrophe-tolerant cloze matching.
- Added recognition cards for 42 routines and 20 modal-particle frames.
- Added Kaikki lemma glosses and Wiktionary form-of redirects.
- Wired the full local pipeline and fixed the Tatoeba links URL.
- Added cloze, gloss, pipeline, and deck checks.
v0.1 — evidence-sized German deck (3,425 cards)
First release of german-pareto-deck.
3,425 cards = 2,963 word cards (1,507 core + 1,456 extension) + 461 pattern cards.
- Words sized by measured coverage curves on two corpora; the 4,000-form cutoff targets the ~95% spoken-comprehension line (Laufer 1989; Adolphs & Schmitt 2003; Nation 2006).
- Patterns: 500 selected under collocation significance (t >= 2, Evert 2004) with PHRASE-List-scale literature allocation; spoken-German frames (Perfekt brackets, separable Satzklammer, modals, Funktionsverbgefuege, particles, routines).
- Every card is an authentic Tatoeba sentence (CC BY 2.0 FR); production-direction and cloze formats per Webb 2009.
- Full traceable methodology: docs/METHODOLOGY.md — decision log D1-D8 incl. the no-arbitrary-thresholds selection registry.
Import: Anki -> File -> Import -> german-pareto-deck.apkg