V2

generate_report.py

@@ -140,6 +140,19 @@ def _normalize_keys(flat: dict) -> dict:
     return {k.replace(".rang", ".niveau"): v for k, v in flat.items()}
 
 
+_LOWER_FIELDS = {"accord_desaccord", "avis_dim_final", "decision_finale"}
+
+def _norm_values(flat: dict) -> dict:
+    """Normalise en minuscule les champs textuels dont la casse varie entre extraction et correction."""
+    out = {}
+    for k, v in flat.items():
+        field = k.split(".")[-1]
+        if field in _LOWER_FIELDS and isinstance(v, str):
+            v = v.strip().lower().replace("desaccord", "désaccord")
+        out[k] = v
+    return out
+
+
 def _get_cat(key: str, ptype: str = "") -> str:
     k = key.lower()
     if ptype == "ELEMENTS_PREUVE":
@@ -210,8 +223,8 @@ def compare_extractions(orig_list, corr_list):
                 ep_counters["json_non_parsable"]["dossiers"].add(fichier)
                 continue
 
-            o_flat = _normalize_keys(_flatten(od))
-            c_flat = _normalize_keys(_flatten(cd))
+            o_flat = _norm_values(_normalize_keys(_flatten(od)))
+            c_flat = _norm_values(_normalize_keys(_flatten(cd)))
             all_keys = set(o_flat) | set(c_flat)
 
             for k in all_keys: