Update smart_breed_matcher.py

smart_breed_matcher.py

@@ -136,34 +136,34 @@ class SmartBreedMatcher:
         desc2_embedding = self._get_cached_embedding(breed2_features['description'])
         description_similarity = float(util.pytorch_cos_sim(desc1_embedding, desc2_embedding))
     
-        # 使用新的精細計算方法
-        size_similarity = self._calculate_size_similarity(breed1_features['size'], breed2_features['size'])
-        exercise_similarity = self._calculate_exercise_similarity(breed1_features['exercise'], breed2_features['exercise'])
+        # 使用改進後的尺寸相似度計算
+        size_similarity = self._calculate_size_similarity(
+            breed1_features['size'],
+            breed2_features['size'],
+            self._get_preferred_size_range(breed1_features['description'])
+        )
     
-        # 性格相似度
+        # 其他相似度計算
+        exercise_similarity = self._calculate_exercise_similarity(breed1_features['exercise'], breed2_features['exercise'])
         temp1_embedding = self._get_cached_embedding(breed1_features['temperament'])
         temp2_embedding = self._get_cached_embedding(breed2_features['temperament'])
         temperament_similarity = float(util.pytorch_cos_sim(temp1_embedding, temp2_embedding))
-    
-        # 健康分數相似度
         health_score1 = self._calculate_health_score(breed1_features['breed_name'])
         health_score2 = self._calculate_health_score(breed2_features['breed_name'])
         health_similarity = 1.0 - abs(health_score1 - health_score2)
-    
-        # 噪音水平相似度
         noise_similarity = self._calculate_noise_similarity(
             breed1_features['breed_name'],
             breed2_features['breed_name']
         )
     
-        # 加權計算
+        # 調整權重，增加尺寸的重要性
         weights = {
-            'description': 0.25,
+            'description': 0.20,  # 降低描述權重
             'temperament': 0.20,
-            'exercise': 0.2,
-            'size': 0.05,
-            'health': 0.15,
-            'noise': 0.15
+            'exercise': 0.20,
+            'size': 0.20,        # 顯著提高尺寸權重
+            'health': 0.10,      # 略微降低
+            'noise': 0.10        # 略微降低
         }
     
         final_similarity = (
@@ -240,14 +240,95 @@ class SmartBreedMatcher:
             'scores': {k: round(v, 4) for k, v in scores.items()}
         }
 
-    def _calculate_size_similarity(self, size1: str, size2: str) -> float:
-        size_map = {'Small': 1, 'Medium': 2, 'Large': 3, 'Giant': 4}
-        value1 = size_map.get(size1, 2)  # 預設為 'Medium'
-        value2 = size_map.get(size2, 2)  # 預設為 'Medium'
+    def _get_preferred_size_range(self, description: str) -> tuple:
+        """分析描述文本，確定用戶偏好的尺寸範圍"""
+        description = description.lower()
+        
+        # 定義關鍵詞匹配
+        size_indicators = {
+            'small': ['small', 'tiny', 'little'],
+            'medium': ['medium', 'medium-sized', 'moderate size'],
+            'medium-large': ['medium to large', 'slightly larger', 'medium-large'],
+            'large': ['large', 'big'],
+            'giant': ['giant', 'huge', 'very large']
+        }
+        
+        # 檢測負面提及
+        negative_indicators = {
+            'small': ['not too small', 'not small'],
+            'large': ['not too large', 'not too big', 'not large'],
+            'giant': ['not giant', 'not huge']
+        }
+        
+        # 默認為中型
+        preferred_min = 2  # medium
+        preferred_max = 3  # large
+        
+        # 分析描述中的尺寸偏好
+        for size, keywords in size_indicators.items():
+            for keyword in keywords:
+                if keyword in description:
+                    if size == 'small':
+                        preferred_min, preferred_max = 1, 2
+                    elif size == 'medium':
+                        preferred_min, preferred_max = 2, 2
+                    elif size == 'medium-large':
+                        preferred_min, preferred_max = 2, 3
+                    elif size == 'large':
+                        preferred_min, preferred_max = 3, 3
+                    elif size == 'giant':
+                        preferred_min, preferred_max = 3, 4
+        
+        # 檢查負面提及並調整
+        for size, keywords in negative_indicators.items():
+            for keyword in keywords:
+                if keyword in description:
+                    if size == 'small':
+                        preferred_min = max(2, preferred_min)
+                    elif size == 'large':
+                        preferred_max = min(2, preferred_max)
+                    elif size == 'giant':
+                        preferred_max = min(3, preferred_max)
+        
+        return (preferred_min, preferred_max)
     
-        # 計算相似度
-        size_similarity = 1.0 - abs(value1 - value2) / 3
-        return max(0.0, size_similarity)  # 確保相似度在 [0, 1] 範圍內
+    def _calculate_size_similarity(self, size1: str, size2: str, preferred_range: tuple = None) -> float:
+        """改進的尺寸相似度計算"""
+        # 更細緻的尺寸映射
+        size_map = {
+            'Tiny': 0.5,
+            'Small': 1,
+            'Small-Medium': 1.5,
+            'Medium': 2,
+            'Medium-Large': 2.5,
+            'Large': 3,
+            'Giant': 4
+        }
+        
+        # 獲取數值
+        value1 = size_map.get(size1, 2)
+        value2 = size_map.get(size2, 2)
+        
+        # 基礎相似度計算
+        base_similarity = 1.0 - (abs(value1 - value2) / 3.5)  # 3.5 是最大可能差異
+        
+        # 如果有偏好範圍，進行額外調整
+        if preferred_range:
+            preferred_min, preferred_max = preferred_range
+            
+            # 檢查是否在偏好範圍內
+            in_range = (preferred_min <= value2 <= preferred_max)
+            
+            # 如果不在範圍內，根據距離降低分數
+            if not in_range:
+                distance_to_range = min(
+                    abs(value2 - preferred_min),
+                    abs(value2 - preferred_max)
+                )
+                penalty = distance_to_range * 0.2  # 每單位差異降低20%
+                base_similarity *= (1 - penalty)
+        
+        return max(0.0, min(1.0, base_similarity))  # 確保在 [0, 1] 範圍內
     
     def _calculate_exercise_similarity(self, exercise1: str, exercise2: str) -> float:
         exercise_map = {'Low': 1, 'Moderate': 2, 'High': 3, 'Very High': 4}