docs(i18n): committer scan_translations.py + 3 heuristiques anti-FP (#633/#644) - #647
Merged
Conversation
…633/#644) Scanner de qualite de traduction reference par #640/#644 mais vivant dans aucun arbre. Commit sous docs/investigations/scripts/ (force-add: dir ignore par .gitignore Scripts/, comme les scripts .ps1 existants) pour reproductibilite. 3 heuristiques anti-FP (baseline 22 -> 1 finding): 1. Whitelist loanwords/mnemoniques (LEGIT_TOKENS): Gish gallop, Creepypasta, Whataboutism, Credo quia absurdum + mnemoniques syllogismes (Barbara..Bamalip). Strips avant script-ratio. 2. Exemption cellules courtes (MIN_SCRIPT_CHARS=12). 3. FR-canon cross-check SAFE: supprime seulement si FR source + traduction partagent un loanword/mnemonique connu. La variante "overlap latin large" masquait le clobber #211 -> restreint au set curve. Resultat master 99145fa: 1 finding = Scenarii 6.1.3 title_zh "Johnny Johnny" (0% CJK) = VRAI defaut (duplication; FR canon "Johnny au Pantheon", PT correct "Johnny no Panthon"). Infirme verdict #644 "KEEP" pour cette cellule (follow-up). 5 tests synthetiques valides. REPO configurable (--repo / $ARGUMENTUM_REPO). Relates #633, #640, #644, #211. Co-Authored-By: Claude-Code <noreply@anthropic.com>
This was referenced Jul 3, 2026
Merged
jsboige
added a commit
that referenced
this pull request
Jul 3, 2026
…iduals NO-OP proof (#661) Secondaire dispatch ai-01 (msg kp26j6) — data pendant of visual verdict #140. Rules audit (Cards/Rules/Argumentum Rules - Cards.csv, read-only): - 15 records x 8 languages = 120/120 cells OK (populated + correct script). - 0 empty, 0 contamination residual post-#640 (scanner #647 = 0). - 2 heuristic flags cleared as false positives: Rules_01 ZH (说谎者学校 present, latin inflation from brand "Argumentum"); embromador (canonical PT game-term for baratineur, 77x in Scenarii PT). Primaire (Fallacies EN residuals) = NO-OP, proven empirically: - 0 empty desc_en across 1408 rows; PK 371/607 (cited empty) already populated (120/79 chars). - Scanner #647 TOTAL=0; 0 EN identical-to-FR; 0 EN reading as French. - The #351/#308 drift (40 PK) was resolved in an earlier cycle. Reporting the no-op rather than fabricating work / spending gpt-5.5 on nothing (pattern: verify before concluding, [[matcher-no-match-is-not-content-absent]]). Read-only, 0 CSV write. Base 9c19e51. Co-authored-by: Your <your.email@example.com> Co-authored-by: Claude-Code <noreply@anthropic.com>
jsboige
added a commit
that referenced
this pull request
Jul 3, 2026
…inding (4 datasets) (#667) Secondaire dispatch 29u572 (ai-01). Final pre-release translation-quality sweep. Re-ran scanner #647 on master 27442ad across all 4 datasets (Fallacies 1408, Virtues 223, Scenarii 167, Rules 15 = 1813 rows). Result: TOTAL: 0 finding — 0 drift, 0 MT-contamination residual, 0 EN-identical-to-FR. Expected (audits anterieurs 100%). Confirms the i18n surface is ship-ready at tag time, cumulative from #640 (Rules refonte), #653 (Johnny fix), #218-295 (Virtues), Scenarii 167/167. No gpt-5.5 fill warranted — reporting the 0-finding state per no-fabrication discipline rather than generating work. Data-side pendant of ai-01's verdict #140 (QA verdict on rendered output remains ai-01's lane). Reproducible: python docs/investigations/scripts/2026-07-02-scan-translations.py Relates #647 #640 #653 #134 #140. Base 27442ad. Co-authored-by: Your <your.email@example.com> Co-authored-by: Claude-Code <noreply@anthropic.com>
jsboige
added a commit
that referenced
this pull request
Jul 4, 2026
…e sign-off, 0 finding) (#679) Secondaire of dispatch `lofjtd` (ai-01). Read-only coherence scan of Fallacies taxonomy family/subfamily labels across 7 non-FR languages - the intra-lang consistency axis not covered by scanner #647 (FR-contamination) nor #192 (FR-relative terminology). Result: 0 inconsistencies at every meaningful granularity: - Family_<lang> by (Famille, Sous-Famille): 29 groups, 0 inconsistent - Subfamily_<lang>: 21 groups, 0 inconsistent - Subsubfamily_<lang>: 63 groups, 0 inconsistent - FR Sous-Famille -> consistent Subfamily_<lang>: 21, 0 inconsistent The apparent Famille-level divergence (7/8 families) is intentional sub-family-level localization, not a data-quality defect. Complements #667 (scanner #647 = 0) and #192 on the three i18n axes. Release Fallacies taxonomy surface: coherent + uncontaminated. 0 CSV write, 0 gpt-5.5 call. Read-only. Co-authored-by: Your <your.email@example.com> Co-authored-by: Claude-Code <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
docs(i18n): committer scan_translations.py + 3 heuristiques anti-FP (#633/#644)
TERTIAIRE (dispatch ai-01 msg-…c60w1b).
scan_translations.pyétait référencé par#640/#644 mais ne vivait dans aucun arbre — commit sous
docs/investigations/scripts/(force-add : dir ignorée par.gitignoreScripts/,comme les scripts
.ps1existants) pour rendre la méthodologie reproductible.Les 3 heuristiques anti-FP (baseline 22 → 1 finding)
LEGIT_TOKENS) : Gish gallop, Creepypasta,Whataboutism, Credo quia absurdum + mnémoniques syllogismes (Barbara, Celarent,
Darii… Bamalip). Strips avant le script-ratio scoring → un titre ZH « Celarent 三段论 »
ne lit plus comme « 50% CJK ».
MIN_SCRIPT_CHARS=12) : les titres <12 lettres(noms propres) sont exonérés du seuil
<30%script-ratio inadapté.traduction partagent un loanword/mnémonique connu. La variante naïve « overlap latin
large » masquait à tort le clobber Full retranslation needed for PT Rules CSV (blocked by #183) #211 (cellule
_ru= FR non traduit partageantles mots FR ordinaires comme « baratineur ») → restreint au set curaté.
Résultat sur master
99145fab22 → 1 finding. L'unique résiduel = Scenarii 6.1.3
title_zh = "Johnny Johnny"(0% CJK, 12 chars) = VRAI défaut (duplication ; FR canon = « Johnny au Panthéon »,
PT correct = « Johnny no Panthéon »).
arbitré « PK6.1.3 title_zh = Johnny Johnny = nom propre, miroir FR canon » — mais le FR
canon est « Johnny au Panthéon », pas « Johnny Johnny ». Le scanner confirme
objectivement (reproductible) que c'est un défaut réel. Comment de correction à poster
sur #644. Les 6 autres arbitrages #644 restent valides (vraies FPs : Credo latin, Gish
gallop/Creepypasta/Whataboutism emprunts, Celarent/Camestres mnémoniques latins).
Validation
5 tests synthétiques (clobber FR détecté ✓, loanwords ZH/RU supprimés ✓, English Channel
PT détecté ✓, Celarent ZH supprimé ✓, FR non traduit détecté ✓).
REPOconfigurable via--repo/$ARGUMENTUM_REPO.Impact
mnémoniques (40-60 cellules style) sont signalés à ai-01 pour décision scope.
Relates #633, #640, #644, #211. Base
99145fab.🤖 Generated with Claude Code