Domi31tls 9bd4729048 fix(c8): remove 'grand' from stopwords (was filtering INSEE name) ...

Le mot "grand" en stopword filtrait les noms INSEE valides
comme GRAND, BILLON-GRAND lors du masquage NER. Sur le corpus
audit_30 : 17 fuites du nom "GRAND" dans
trackare-05012965-23060770.

Fix : suppression de la ligne (pipeline INSEE exige contexte
fort pour masquer, "grand" minuscule isolé ne sera pas FP).

Tests à venir : tests/unit/test_c8_grand_regression.py (Qwen)
Ref: docs/coordination/inbox/for-dom/2026-05-29_qwen_analyse-regression-grand.md

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

2026-05-29 17:58:54 +02:00

..

bdpm

Externalize dictionaries and add anonymization review corpus

2026-04-21 10:32:57 +02:00

finess

Externalize dictionaries and add anonymization review corpus

2026-04-21 10:32:57 +02:00

insee

feat: réduction FP + gazetteers adresses FINESS + batch parallèle + corrections multi-axes

2026-03-16 09:26:56 +01:00

quaero

feat(phase2): Fine-tuning CamemBERT-bio v2 (F1=0.90) + enrichissement données

2026-03-10 02:06:08 +01:00

silver_annotations

feat: réduction FP + gazetteers adresses FINESS + batch parallèle + corrections multi-axes

2026-03-16 09:26:56 +01:00

companion_blacklist.txt

refactor: externaliser DPI labels et companion blacklist (modifiables sans recompiler)

2026-04-14 10:26:18 +02:00

dpi_labels_blacklist.txt

refactor: externaliser DPI labels et companion blacklist (modifiables sans recompiler)

2026-04-14 10:26:18 +02:00

stopwords_manuels.txt

fix(c8): remove 'grand' from stopwords (was filtering INSEE name)

2026-05-29 17:58:54 +02:00

villes_blacklist.txt

feat: externalisation des listes — stop-words et villes modifiables sans code

2026-03-31 07:45:42 +02:00