Update column_resolver.py

column_resolver.py

@@ -1,18 +1,29 @@
 import re
-from typing import List, Dict
+from typing import List, Dict, Tuple
+
+ROLE_SYNONYMS = {
+    "facility": ["facility", "hospital", "site", "centre", "clinic", "settlement", "community"],
+    "zone": ["zone", "region", "area", "district"],
+    "specialty": ["specialty", "service", "discipline"],
+    "population": ["population", "residents", "members"],
+    "cost": ["cost", "expense", "spend"],
+    "outcome": ["outcome", "improvement", "result"]
+}
 
 def _norm(s: str) -> str:
     return re.sub(r'[^a-z0-9]+', '', s.lower())
 
-def build_alias_map(columns: List[str]) -> Dict[str, str]:
-    m = {}
+def resolve_role(role: str, columns: List[str]) -> Tuple[str | None, float]:
+    """Return best matching column for semantic role with confidence score."""
+    target = _norm(role)
+    syns = ROLE_SYNONYMS.get(target, [role])
+    best, score = None, 0.0
     for c in columns:
-        n = _norm(c)
-        m[n] = c
-        if n.endswith('s') and n[:-1] not in m: m[n[:-1]] = c
-        if not n.endswith('s') and n+'s' not in m: m[n+'s'] = c
-    return m
-
-def resolve_cols(requested: List[str], available: List[str]) -> List[str]:
-    alias = build_alias_map(available)
-    return [alias.get(_norm(r), r) for r in requested]
+        nc = _norm(c)
+        for s in syns:
+            ns = _norm(s)
+            if ns in nc or nc in ns:
+                sc = len(ns) / max(len(nc), 1)
+                if sc > score:
+                    best, score = c, sc
+    return best, score