Dom/anonymisation
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix(phase2): Ajout stop words cliniques — 117 FP en moins (RESPI, NEPHRO, URINE, etc.)

Browse Source 
Termes cliniques Trackare (RESPI, NEPHRO, CARDIO, PULMO, POST-OP, SPO2, etc.)
et termes médicaux (respiratoire, rénale, cardiaque, urine) ajoutés aux stop words.
Filtrés par NER EDS-Pseudo et selective_rescan. 0 fuite, 0 régression.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Domi31tls
2026-03-09 09:58:58 +01:00
parent 8629a0cda0
commit 782551c1c6
12 changed files with 116 additions and 108 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
anonymizer_core_refactored_onnx.py
View File

@@ -466,6 +466,14 @@ _MEDICAL_STOP_WORDS_SET = {
"morphine", "claforan", "skenan", "actiskenan",
# Fragments de noms de médicaments (pdfplumber split)
"sium", "pegic", "fenid", "profenid",
# Catégories cliniques Trackare (en-têtes de section masqués à tort)
"respi", "respiratoire", "nephro", "cardio", "neuro", "onco", "pulmo",
"hemato", "hémato", "infectieux", "thermie", "diurese", "diurèse",
"transit", "anemie", "anémie", "constantes", "examen",
"post-op", "postop", "pré-op", "preop", "chimio", "elim",
"toilette", "sommeil", "hypota", "hypotension", "spo2",
"urine", "urines", "sng",
"rénale", "renale", "rénal", "renal", "cardiaque",
# Termes structurels trackare
"transmissions", "transmission", "releve", "relevé",
"objectif", "objectifs", "evaluation", "évaluation",