diff --git a/graphify/build.py b/graphify/build.py
index 68f50bc9f..582f3f2ac 100644
--- a/graphify/build.py
+++ b/graphify/build.py
@@ -228,8 +228,9 @@ def build(
 
 
 def _norm_label(label: str) -> str:
-    """Canonical dedup key — lowercase, alphanumeric only."""
-    return re.sub(r"[^a-z0-9 ]", "", label.lower()).strip()
+    """Canonical dedup key — Unicode-aware, preserves CJK/word characters."""
+    label = unicodedata.normalize("NFKC", label)
+    return re.sub(r"[\W_ ]+", " ", label.casefold(), flags=re.UNICODE).strip()
 
 
 def deduplicate_by_label(nodes: list[dict], edges: list[dict]) -> tuple[list[dict], list[dict]]:
diff --git a/graphify/dedup.py b/graphify/dedup.py
index 5c15f33f5..dc177fb8f 100644
--- a/graphify/dedup.py
+++ b/graphify/dedup.py
@@ -6,6 +6,7 @@ Jaro-Winkler verification → same-community boost → union-find merge.
 from __future__ import annotations
 import math
 import re
+import unicodedata
 from collections import defaultdict
 
 from datasketch import MinHash, MinHashLSH
@@ -15,8 +16,9 @@ from rapidfuzz.distance import JaroWinkler
 # ── helpers ───────────────────────────────────────────────────────────────────
 
 def _norm(label: str) -> str:
-    """Lowercase + collapse non-alphanumeric runs to space."""
-    return re.sub(r"[^a-z0-9]+", " ", label.lower()).strip()
+    """Lowercase + collapse non-alphanumeric runs to space (Unicode-aware)."""
+    label = unicodedata.normalize("NFKC", label)
+    return re.sub(r"[\W_]+", " ", label.casefold(), flags=re.UNICODE).strip()
 
 
 def _entropy(label: str) -> float: