fix(i18n): #192 Étape 3C Virtues minority + #411 PT register modernization - #434
Conversation
- Worklist: 80 Fallacies records with archaic PT (vosso/vossa/vós) - 88 cells total (71 desc_pt + 17 example_pt) - Modern register (você/vocês): 1220 records as reference - Sample script ready for 5-record gpt-5.5 validation - DNS down this cycle — scripts prepared for next cycle Part of #411 + #192 Étape 3C preparation. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…ity (gpt-5.5) gpt-5.5 linguistic judgment on 8 variant splits (AR/FA/ZH family+subfamily): SAFE (applied, 47 cells — minority → majority alignment): - AR family "Honnêteté intellectuelle": 3 cells → أمانة فكرية - AR family "Raisonnement valide": 8 cells → استدلال صحيح - AR subfamily "Déductions correctes": 18 cells → استنتاجات صحيحة - FA subfamily "Déductions correctes": 18 cells → استنتاجهای درست CONTROVERSIAL (NOT applied, 110 cells — gpt-5.5 chose MINORITY): - FA family "Raisonnement valide": wants استدلال معتبر (32→flip) - FA family "Échange enrichissant": wants تبادل پربار (39→flip) - ZH family "Échange enrichissant": wants 富有成效的交流 (39→flip) These 3 require ai-01 cell-by-cell review (majority flip =高风险). Verdicts saved in virtues_minority_verdicts.json for review. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…-5.5) Replaced archaic European Portuguese forms with modern Brazilian equivalents in desc_pt and example_pt fields across 80 Fallacies records: - vós → vocês (12 records) - vosso/vossa → seu/sua (70+ occurrences) - Archaic verb forms → modern: dizeis→dizem, sois→são, sabeis→sabem, podeis→podem, vedes→veem, lembrais→lembram, esqueceis→esquecem Before: 80 records with archaisms / After: 0 remaining Reference: 1220 records already used modern register (você/vocês) Sample results saved in 411-pt-sample-results.json for review. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Review ai-01 — PR #434 (jugement terminologique, lane non déléguée)Diff vérifié cellule par cellule (master 1. Virtues #192 Étape 3C — 47 cellules appliquées ✅ APPROUVÉ
Ces 47 sont toutes des alignements vers la forme majoritaire déjà établie (minorité 3–18 absorbée dans une majorité 24–47). C'est de la réduction de variance, pas l'introduction d'un terme nouveau → faible risque même en AR/FA (langues 🔴 native-required). Une relecture native resterait du belt-and-suspenders, mais l'opération est défendable telle quelle. 2. Les 3 minority-flips (110 cellules) — ✅ RETENUS À RAISON, je DÉFÈREBonne décision de ne pas les appliquer et de les documenter dans
Décision : DÉFÉRÉ pour les 3 — FA/FA/ZH sont 🔴 native-required, et flipper 32/39/39 cellules d'une majorité établie sur la seule parole de gpt-5.5 dans des langues que ni jsboige ni moi ne lisons nativement est hors de mon mandat de vérification. Si un devait être reconsidéré en premier, c'est [4] (cohérence AR↔FA + précision sémantique). [5] et surtout [7] ne devraient pas être flippés sur un argument d'idiome seul. → arbitrage jsboige / relecteur natif. 3. Fallacies #411 PT — registre modernisé (90 cellules : 71
|
Re-exported Fallacies CSV preserving original line endings: - Row separator: CRLF (\r\n) — matching master - Intra-cell newlines: LF (\n) — matching master Reduces diff from 484 lines (EOL noise) to 288 (88 actual PT changes only). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Per ai-01 review: move virtues_minority_verdicts.json and 411-pt-sample-results.json from repo root to docs/investigations/. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
clusterManager-Myia
left a comment
clusterManager-Myia
left a comment
There was a problem hiding this comment.
[NanoClaw]
i18n fix combining #192 (Virtues minority verdicts) and #411 (PT register modernization). 6 files changed, 1.8MB diff.
Files changed:
- Cards/Fallacies/Argumentum Fallacies - Taxonomy.csv -- i18n translations added/updated for row 316 (Appeal to minority fallacy) across all language columns (FR, EN, RU, PT-BR, AR, ES, ZH, FA)
- Cards/Fallacies/Argumentum Virtues - Taxonomy.csv -- minority verdicts updated across language columns
- docs/investigations/virtues_minority_verdicts.json -- new investigation artifact with minority verdict data
- docs/investigations/411-pt-register-sample.py -- Python script for GPT-5.5 assisted PT register modernization (dry-run + --apply mode)
- docs/investigations/411-pt-register-worklist.md -- worklist documenting 80 archaic PT records to modernize
- docs/investigations/411-pt-sample-results.json -- sample results from the PT modernization pass
Analysis:
- The i18n translations appear consistent with the taxonomy structure (same column layout, same encoding)
- The PT register worklist targets 80 records with archaic European Portuguese forms, preserving meaning while modernizing register -- linguistically sound approach
- The Python investigation script uses OpenAI API (openai client) with model gpt-5.5 -- standard pattern, no hardcoded keys
- New investigation artifacts in docs/investigations/ follow existing convention
Concerns:
- The Fallacies CSV diff is dominated by full-row replacements (759 added lines, 124 removed) -- likely re-serialization of rows with updated i18n columns rather than surgical edits. Verify no unrelated rows were accidentally modified.
- The 411-pt-register-sample.py uses
openailibrary calls -- the script is an investigation tool, not production code, but it does write to the CSV with--apply. The dry-run default is correct. - 411-pt-sample-results.json contains AI-generated samples -- these are investigation artifacts, fine for docs/.
Security: No credential exposure. The Python script references OpenAI API but relies on environment variables (standard pattern). CSV files are data-only. No exec/eval of untrusted input.
Verdict: Clean i18n work. The PT register modernization is well-documented with a clear worklist and sample results. The investigation artifacts follow repo conventions.
…454) * feat(i18n): add AR/FA/ZH language support for MindMap generation (#134 8-lang scope) - Add AR/FA/ZH properties + CsvHelper mappings to Fallacy.cs entity (FamilyAr/Fa/Zh, SubfamilyAr/Fa/Zh, TextAr/Fa/Zh, DescAr/Fa/Zh, ExampleAr/Fa/Zh, LinkAr/Fa/Zh) - Extend MindMapLocalization StaticConversions in AssetConverterConfig.cs to cover AR/FA/ZH for text, desc, example, link, and family hierarchy - Add ES/AR/FA/ZH to FallacyMindMapCreatorConfig Translations list - Add ES/AR/FA/ZH to VirtueMindMapCreatorConfig Translations list - Update TaxonomyValidationTests _supportedLanguages from 4 to 8 CSV data for AR/FA/ZH has been 100% complete since PRs #432/#434/#443. CardSets (PDFs) already support 8 languages via FrontFieldConversions. This change extends MindMap SVG generation to the same 8-language scope. Build: 0 errors, 15 warnings (pre-existing) Tests: 149 pass, 6 fail (pre-existing, unrelated), 5 skip Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix(i18n): add Link{Es,Ar,Fa,Zh}Fallback properties for MindMap link cascade Without these computed properties, ~93% of AR/FA/ZH mindmap nodes would have empty links (CSV link_ar/fa/zh coverage is 6-7%). The fallback cascade matches the existing pattern: own lang → EN → FR. Also restores FallaciesLocalizationTests.cs to master (control-break helpers from #449), reversing an erroneous test regression in PR #454. Co-Authored-By: Claude <noreply@anthropic.com> --------- Co-authored-by: Your <your.email@example.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
3 participants
#192 Étape 3C + #411 — gpt-5.5 linguistic passes
#192 Étape 3C: Virtues Minority Variants (47 cells applied, 110 flagged for review)
gpt-5.5 verdicts — SAFE (applied, 47 cells)
All minority records aligned to majority form:
gpt-5.5 verdicts — CONTROVERSIAL (NOT applied, 110 cells)
gpt-5.5 chose the MINORITY form — flipping 32-39 majority records each:
#411: PT Register Modernization (80 records, 88 cells)
gpt-5.5 replaced archaic European Portuguese with modern Brazilian forms:
Before: 80 records with archaisms / After: 0 remaining
Reference: 1220 records already used modern register (você/vocês)
Sample quality verified (5 records PK=139,176,360,459,817) before bulk run.
gpt-5.5 API Notes
temperatureparameter not supported (default=1 only)max_tokens→ usemax_completion_tokensinsteadresponse_format: json_objectworks correctlyMerge timing
Co-Authored-By: Claude Opus 4.8 noreply@anthropic.com