Skip to content

data(virtues): translate Virtues to ES (1401/1561 cells, 90%) - #363

Merged
jsboige merged 1 commit into
masterfrom
data/virtues-es-gpt55
May 27, 2026
Merged

data(virtues): translate Virtues to ES (1401/1561 cells, 90%)#363
jsboige merged 1 commit into
masterfrom
data/virtues-es-gpt55

Conversation

@jsboige

@jsboige jsboige commented May 27, 2026

Copy link
Copy Markdown
Contributor

Summary

  • Translate Virtues taxonomy to Spanish (ES) using DatasetUpdater with gpt-5.5
  • 1401/1561 target cells filled (90%) across 7 ES fields for 223 Virtues entries
  • Add ES translation prompts (user + assistant) for DatasetUpdater
  • Add Virtues ES task config (disabled after run)

Coverage by field

Field Filled Total
title_es 220 223
description_es 220 223
remark_es 220 223
family_es 220 223
subfamily_es 212 223
subsubfamily_es 192 223
link_es 117 223

Quality checks performed

  • Spot-checked 15 random entries across all 8 families — translations are idiomatic, use correct declarative register
  • Canonical family mappings respected (Argumento pertinente, Presentación íntegra, etc.)
  • 4 inconsistent subfamily/subsubfamily translations normalized to most frequent variant
  • No FR contamination detected

Remaining work

  • 160 cells still empty (31 subsubfamily_es, 11 subfamily_es, 3 rows missing entirely, 106 link_es)
  • Pass 2 needed after forensic analysis of API call behavior (runaway bug investigation in progress)

⚠️ Known issue

Pass 1 made significantly more API calls than expected (~3700+ for 1561 cells). The DatasetUpdater's SelectEmptyTargets + chunking behavior needs forensic analysis before pass 2. All tasks are disabled and Mode is restored to pipeline defaults.

🤖 Generated with Claude Code

- Fill ES fields for 220/223 Virtues entries via DatasetUpdater (gpt-5.5)
- Coverage: title_es 220/223, description_es 220/223, remark_es 220/223,
  family_es 220/223, subfamily_es 212/223, subsubfamily_es 192/223,
  link_es 117/223
- Normalize 4 inconsistent subfamily/subsubfamily translations
- Add ES translation prompts (user + assistant)
- Add DatasetUpdater task config for Virtues ES (disabled after run)
- Disable Fallacies cleanup task (leftover from PR #362)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

@clusterManager-Myia clusterManager-Myia left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[Hermes] — COMMENT_WITH_CONCERNS

PR: data(virtues): translate Virtues to ES (1401/1561 cells, 90%)
SHA: 171de38

Checklist

Check Status
Security scan ✅ CLEAN
Cross-repo impact ✅ N/A (data CSV + pipeline config)
CI ✅ 3/3 pass (GitGuardian, build Debug, build Release)
Files changed 4 files (1 CSV + 2 prompt templates + 1 C# config), +328/-222

Concerns — 2 items (0 bloquants)

1. [NON-BLOQUANT] Coverage 90% — 10% non traduit

Le titre indique 1401/1561 cells (90%). Les ~160 cells restantes non traduites correspondent probablement aux entrées sans description_es ou remark_es dans les colonnes finales du CSV. Ce n'est pas un bug (le pipeline SelectEmptyTargets = true ne traduit que les vides), mais il faudra un suivi pour les 10% restants.

Suggestion : Ouvrir un ticket de suivi pour la complétion des 10% restants, ou préciser dans la PR description pourquoi ces entrées sont volontairement non traduites.

2. [NON-BLOQUANT] Pipeline cleanup Fallacies Enabled = true → false

Le changement Enabled = true → false sur le pipeline Cleanup Fallacies est un side-effect de cette PR. Normalement ce genre de changement de config devrait être dans une PR séparée. Cependant, c'est un one-liner dans un fichier de config, donc le risque est minimal.

Verdict : COMMENT_WITH_CONCERNS (pas de blocage)

Traduction ES de qualité : registre neutre, cohérence taxonomique respectée, prompts bien structurés avec function calling. Le pipeline C# est correctement configuré.

@jsboige
jsboige merged commit 180cad5 into master May 27, 2026
3 checks passed
@jsboige
jsboige deleted the data/virtues-es-gpt55 branch May 27, 2026 10:04
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants