Create fusion.py

core/fusion.py

@@ -0,0 +1,22 @@
+from typing import Dict
+
+# judge_score is 1..5 (int). nlp_subscore is 0..1. alpha in [0,1]
+
+def fuse_metric(judge_score, nlp_subscore: float, alpha: float) -> float:
+    js = 0.0
+    if judge_score is not None:
+        try:
+            js = max(0.0, min(1.0, float(judge_score) / 5.0))
+        except Exception:
+            js = 0.0
+    ns = max(0.0, min(1.0, float(nlp_subscore)))
+    fused_0_1 = alpha * js + (1 - alpha) * ns
+    return round(fused_0_1 * 10.0, 2)
+
+
+def weighted_total(metric_scores_0_10: Dict[str, float], weights: Dict[str, float]) -> float:
+    tot = 0.0
+    for k, v in metric_scores_0_10.items():
+        w = weights.get(k, 0.0)
+        tot += (v or 0.0) * w
+    return round(tot, 2)