Spaces:
Running
on
Zero
Running
on
Zero
Delete smart_breed_matcher.py
Browse files- smart_breed_matcher.py +0 -376
smart_breed_matcher.py
DELETED
@@ -1,376 +0,0 @@
|
|
1 |
-
import torch
|
2 |
-
import numpy as np
|
3 |
-
from typing import List, Dict, Tuple, Optional
|
4 |
-
from dataclasses import dataclass
|
5 |
-
from breed_health_info import breed_health_info
|
6 |
-
from breed_noise_info import breed_noise_info
|
7 |
-
from dog_database import dog_data
|
8 |
-
from scoring_calculation_system import UserPreferences
|
9 |
-
from sentence_transformers import SentenceTransformer, util
|
10 |
-
|
11 |
-
class SmartBreedMatcher:
|
12 |
-
def __init__(self, dog_data: List[Tuple]):
|
13 |
-
self.dog_data = dog_data
|
14 |
-
self.model = SentenceTransformer('paraphrase-MiniLM-L6-v2')
|
15 |
-
|
16 |
-
def _categorize_breeds(self) -> Dict:
|
17 |
-
"""自動將狗品種分類"""
|
18 |
-
categories = {
|
19 |
-
'working_dogs': [],
|
20 |
-
'herding_dogs': [],
|
21 |
-
'hunting_dogs': [],
|
22 |
-
'companion_dogs': [],
|
23 |
-
'guard_dogs': []
|
24 |
-
}
|
25 |
-
|
26 |
-
for breed_info in self.dog_data:
|
27 |
-
description = breed_info[9].lower()
|
28 |
-
temperament = breed_info[4].lower()
|
29 |
-
|
30 |
-
# 根據描述和性格特徵自動分類
|
31 |
-
if any(word in description for word in ['herding', 'shepherd', 'cattle', 'flock']):
|
32 |
-
categories['herding_dogs'].append(breed_info[1])
|
33 |
-
elif any(word in description for word in ['hunting', 'hunt', 'retriever', 'pointer']):
|
34 |
-
categories['hunting_dogs'].append(breed_info[1])
|
35 |
-
elif any(word in description for word in ['companion', 'toy', 'family', 'lap']):
|
36 |
-
categories['companion_dogs'].append(breed_info[1])
|
37 |
-
elif any(word in description for word in ['guard', 'protection', 'watchdog']):
|
38 |
-
categories['guard_dogs'].append(breed_info[1])
|
39 |
-
elif any(word in description for word in ['working', 'draft', 'cart']):
|
40 |
-
categories['working_dogs'].append(breed_info[1])
|
41 |
-
|
42 |
-
return categories
|
43 |
-
|
44 |
-
def find_similar_breeds(self, breed_name: str, top_n: int = 5) -> List[Tuple[str, float]]:
|
45 |
-
"""找出與指定品種最相似的其他品種"""
|
46 |
-
target_breed = next((breed for breed in self.dog_data if breed[1] == breed_name), None)
|
47 |
-
if not target_breed:
|
48 |
-
return []
|
49 |
-
|
50 |
-
# 獲取目標品種的特徵
|
51 |
-
target_features = {
|
52 |
-
'breed_name': target_breed[1], # 添加品種名稱
|
53 |
-
'size': target_breed[2],
|
54 |
-
'temperament': target_breed[4],
|
55 |
-
'exercise': target_breed[7],
|
56 |
-
'description': target_breed[9]
|
57 |
-
}
|
58 |
-
|
59 |
-
similarities = []
|
60 |
-
for breed in self.dog_data:
|
61 |
-
if breed[1] != breed_name:
|
62 |
-
breed_features = {
|
63 |
-
'breed_name': breed[1], # 添加品種名稱
|
64 |
-
'size': breed[2],
|
65 |
-
'temperament': breed[4],
|
66 |
-
'exercise': breed[7],
|
67 |
-
'description': breed[9]
|
68 |
-
}
|
69 |
-
|
70 |
-
similarity_score = self._calculate_breed_similarity(target_features, breed_features)
|
71 |
-
similarities.append((breed[1], similarity_score))
|
72 |
-
|
73 |
-
return sorted(similarities, key=lambda x: x[1], reverse=True)[:top_n]
|
74 |
-
|
75 |
-
|
76 |
-
def _calculate_breed_similarity(self, breed1_features: Dict, breed2_features: Dict) -> float:
|
77 |
-
"""計算兩個品種之間的相似度,包含健康和噪音因素"""
|
78 |
-
# 計算描述文本的相似度
|
79 |
-
desc1_embedding = self.model.encode(breed1_features['description'])
|
80 |
-
desc2_embedding = self.model.encode(breed2_features['description'])
|
81 |
-
description_similarity = float(util.pytorch_cos_sim(desc1_embedding, desc2_embedding))
|
82 |
-
|
83 |
-
# 基本特徵相似度
|
84 |
-
size_similarity = 1.0 if breed1_features['size'] == breed2_features['size'] else 0.5
|
85 |
-
exercise_similarity = 1.0 if breed1_features['exercise'] == breed2_features['exercise'] else 0.5
|
86 |
-
|
87 |
-
# 性格相似度
|
88 |
-
temp1_embedding = self.model.encode(breed1_features['temperament'])
|
89 |
-
temp2_embedding = self.model.encode(breed2_features['temperament'])
|
90 |
-
temperament_similarity = float(util.pytorch_cos_sim(temp1_embedding, temp2_embedding))
|
91 |
-
|
92 |
-
# 健康分數相似度
|
93 |
-
health_score1 = self._calculate_health_score(breed1_features['breed_name'])
|
94 |
-
health_score2 = self._calculate_health_score(breed2_features['breed_name'])
|
95 |
-
health_similarity = 1.0 - abs(health_score1 - health_score2)
|
96 |
-
|
97 |
-
# 噪音水平相似度
|
98 |
-
noise_similarity = self._calculate_noise_similarity(
|
99 |
-
breed1_features['breed_name'],
|
100 |
-
breed2_features['breed_name']
|
101 |
-
)
|
102 |
-
|
103 |
-
# 加權計算
|
104 |
-
weights = {
|
105 |
-
'description': 0.25,
|
106 |
-
'temperament': 0.20,
|
107 |
-
'exercise': 0.15,
|
108 |
-
'size': 0.10,
|
109 |
-
'health': 0.15,
|
110 |
-
'noise': 0.15
|
111 |
-
}
|
112 |
-
|
113 |
-
final_similarity = (
|
114 |
-
description_similarity * weights['description'] +
|
115 |
-
temperament_similarity * weights['temperament'] +
|
116 |
-
exercise_similarity * weights['exercise'] +
|
117 |
-
size_similarity * weights['size'] +
|
118 |
-
health_similarity * weights['health'] +
|
119 |
-
noise_similarity * weights['noise']
|
120 |
-
)
|
121 |
-
|
122 |
-
return final_similarity
|
123 |
-
|
124 |
-
|
125 |
-
def _calculate_final_scores(self, breed_name: str, base_scores: Dict,
|
126 |
-
smart_score: float, is_preferred: bool,
|
127 |
-
similarity_score: float = 0.0) -> Dict:
|
128 |
-
"""
|
129 |
-
計算最終分數,包含基礎分數和獎勵分數
|
130 |
-
|
131 |
-
Args:
|
132 |
-
breed_name: 品種名稱
|
133 |
-
base_scores: 基礎評分 (空間、運動等)
|
134 |
-
smart_score: 智能匹配分數
|
135 |
-
is_preferred: 是否為用戶指定品種
|
136 |
-
similarity_score: 與指定品種的相似度 (0-1)
|
137 |
-
"""
|
138 |
-
# 基礎權重
|
139 |
-
weights = {
|
140 |
-
'base': 0.6, # 基礎分數權重
|
141 |
-
'smart': 0.25, # 智能匹配權重
|
142 |
-
'bonus': 0.15 # 獎勵分數權重
|
143 |
-
}
|
144 |
-
|
145 |
-
# 計算基礎分數
|
146 |
-
base_score = base_scores.get('overall', 0.7)
|
147 |
-
|
148 |
-
# 計算獎勵分數
|
149 |
-
bonus_score = 0.0
|
150 |
-
if is_preferred:
|
151 |
-
# 用戶指定品種獲得最高獎勵
|
152 |
-
bonus_score = 0.95
|
153 |
-
elif similarity_score > 0:
|
154 |
-
# 相似品種獲得部分獎勵,但不超過80%的最高獎勵
|
155 |
-
bonus_score = min(0.8, similarity_score) * 0.95
|
156 |
-
|
157 |
-
# 計算最終分數
|
158 |
-
final_score = (
|
159 |
-
base_score * weights['base'] +
|
160 |
-
smart_score * weights['smart'] +
|
161 |
-
bonus_score * weights['bonus']
|
162 |
-
)
|
163 |
-
|
164 |
-
# 更新各項分數
|
165 |
-
scores = base_scores.copy()
|
166 |
-
|
167 |
-
# 如果是用戶指定品種,稍微提升各項基礎分數,但保持合理範圍
|
168 |
-
if is_preferred:
|
169 |
-
for key in scores:
|
170 |
-
if key != 'overall':
|
171 |
-
scores[key] = min(1.0, scores[key] * 1.1) # 最多提升10%
|
172 |
-
|
173 |
-
# 為相似品種調整分數
|
174 |
-
elif similarity_score > 0:
|
175 |
-
boost_factor = 1.0 + (similarity_score * 0.05) # 最多提升5%
|
176 |
-
for key in scores:
|
177 |
-
if key != 'overall':
|
178 |
-
scores[key] = min(0.95, scores[key] * boost_factor) # 確保不超過95%
|
179 |
-
|
180 |
-
return {
|
181 |
-
'final_score': round(final_score, 4),
|
182 |
-
'base_score': round(base_score, 4),
|
183 |
-
'bonus_score': round(bonus_score, 4),
|
184 |
-
'scores': {k: round(v, 4) for k, v in scores.items()}
|
185 |
-
}
|
186 |
-
|
187 |
-
def _calculate_health_score(self, breed_name: str) -> float:
|
188 |
-
"""計算品種的健康分數"""
|
189 |
-
if breed_name not in breed_health_info:
|
190 |
-
return 0.5
|
191 |
-
|
192 |
-
health_notes = breed_health_info[breed_name]['health_notes'].lower()
|
193 |
-
|
194 |
-
# 嚴重健康問題
|
195 |
-
severe_conditions = [
|
196 |
-
'cancer', 'cardiomyopathy', 'epilepsy', 'dysplasia',
|
197 |
-
'bloat', 'progressive', 'syndrome'
|
198 |
-
]
|
199 |
-
|
200 |
-
# 中等健康問題
|
201 |
-
moderate_conditions = [
|
202 |
-
'allergies', 'infections', 'thyroid', 'luxation',
|
203 |
-
'skin problems', 'ear'
|
204 |
-
]
|
205 |
-
|
206 |
-
severe_count = sum(1 for condition in severe_conditions if condition in health_notes)
|
207 |
-
moderate_count = sum(1 for condition in moderate_conditions if condition in health_notes)
|
208 |
-
|
209 |
-
health_score = 1.0
|
210 |
-
health_score -= (severe_count * 0.1)
|
211 |
-
health_score -= (moderate_count * 0.05)
|
212 |
-
|
213 |
-
# 特殊條件調整(根據用戶偏好)
|
214 |
-
if hasattr(self, 'user_preferences'):
|
215 |
-
if self.user_preferences.has_children:
|
216 |
-
if 'requires frequent' in health_notes or 'regular monitoring' in health_notes:
|
217 |
-
health_score *= 0.9
|
218 |
-
|
219 |
-
if self.user_preferences.health_sensitivity == 'high':
|
220 |
-
health_score *= 0.9
|
221 |
-
|
222 |
-
return max(0.3, min(1.0, health_score))
|
223 |
-
|
224 |
-
|
225 |
-
|
226 |
-
def _calculate_noise_similarity(self, breed1: str, breed2: str) -> float:
|
227 |
-
"""計算兩個品種的噪音相似度"""
|
228 |
-
noise_levels = {
|
229 |
-
'Low': 1,
|
230 |
-
'Moderate': 2,
|
231 |
-
'High': 3,
|
232 |
-
'Unknown': 2 # 默認為中等
|
233 |
-
}
|
234 |
-
|
235 |
-
noise1 = breed_noise_info.get(breed1, {}).get('noise_level', 'Unknown')
|
236 |
-
noise2 = breed_noise_info.get(breed2, {}).get('noise_level', 'Unknown')
|
237 |
-
|
238 |
-
# 獲取數值級別
|
239 |
-
level1 = noise_levels.get(noise1, 2)
|
240 |
-
level2 = noise_levels.get(noise2, 2)
|
241 |
-
|
242 |
-
# 計算差異並歸一化
|
243 |
-
difference = abs(level1 - level2)
|
244 |
-
similarity = 1.0 - (difference / 2) # 最大差異是2,所以除以2來歸一化
|
245 |
-
|
246 |
-
return similarity
|
247 |
-
|
248 |
-
def _general_matching(self, description: str, top_n: int = 10) -> List[Dict]:
|
249 |
-
"""基本的品種匹配邏輯,考慮描述、性格、噪音和健康因素"""
|
250 |
-
matches = []
|
251 |
-
for breed in self.dog_data:
|
252 |
-
breed_name = breed[1]
|
253 |
-
breed_description = breed[9]
|
254 |
-
temperament = breed[4]
|
255 |
-
|
256 |
-
# 計算描述文本和性格的相似度
|
257 |
-
desc_embedding = self.model.encode(description)
|
258 |
-
breed_desc_embedding = self.model.encode(breed_description)
|
259 |
-
breed_temp_embedding = self.model.encode(temperament)
|
260 |
-
|
261 |
-
desc_similarity = float(util.pytorch_cos_sim(desc_embedding, breed_desc_embedding))
|
262 |
-
temp_similarity = float(util.pytorch_cos_sim(desc_embedding, breed_temp_embedding))
|
263 |
-
|
264 |
-
# 計算噪音相似度和健康分數
|
265 |
-
noise_similarity = self._calculate_noise_similarity(breed_name, breed_name)
|
266 |
-
health_score = self._calculate_health_score(breed_name)
|
267 |
-
health_similarity = 1.0 - abs(health_score - 0.8) # 假設理想健康分數為 0.8
|
268 |
-
|
269 |
-
# 加權計算分數
|
270 |
-
weights = {
|
271 |
-
'description': 0.35,
|
272 |
-
'temperament': 0.25,
|
273 |
-
'noise': 0.2,
|
274 |
-
'health': 0.2
|
275 |
-
}
|
276 |
-
|
277 |
-
# 計算最終分數
|
278 |
-
final_score = (
|
279 |
-
desc_similarity * weights['description'] +
|
280 |
-
temp_similarity * weights['temperament'] +
|
281 |
-
noise_similarity * weights['noise'] +
|
282 |
-
health_similarity * weights['health']
|
283 |
-
)
|
284 |
-
|
285 |
-
matches.append({
|
286 |
-
'breed': breed_name,
|
287 |
-
'score': final_score,
|
288 |
-
'is_preferred': False,
|
289 |
-
'similarity': final_score,
|
290 |
-
'reason': "Matched based on description, temperament, noise level, and health score"
|
291 |
-
})
|
292 |
-
|
293 |
-
# 排序並返回前 N 個匹配結果
|
294 |
-
return sorted(matches, key=lambda x: -x['score'])[:top_n]
|
295 |
-
|
296 |
-
|
297 |
-
def match_user_preference(self, description: str, top_n: int = 10) -> List[Dict]:
|
298 |
-
"""根據用戶描述匹配最適合的品種"""
|
299 |
-
preferred_breed = self._detect_breed_preference(description)
|
300 |
-
|
301 |
-
matches = []
|
302 |
-
if preferred_breed:
|
303 |
-
similar_breeds = self.find_similar_breeds(preferred_breed, top_n=top_n)
|
304 |
-
|
305 |
-
# 首先添加偏好品種
|
306 |
-
breed_info = next((breed for breed in self.dog_data if breed[1] == preferred_breed), None)
|
307 |
-
if breed_info:
|
308 |
-
health_score = self._calculate_health_score(preferred_breed)
|
309 |
-
noise_info = breed_noise_info.get(preferred_breed, {
|
310 |
-
"noise_level": "Unknown",
|
311 |
-
"noise_notes": "No noise information available"
|
312 |
-
})
|
313 |
-
|
314 |
-
# 偏好品種必定是最高分
|
315 |
-
matches.append({
|
316 |
-
'breed': preferred_breed,
|
317 |
-
'score': 1.0,
|
318 |
-
'is_preferred': True,
|
319 |
-
'similarity': 1.0,
|
320 |
-
'health_score': health_score,
|
321 |
-
'noise_level': noise_info['noise_level'],
|
322 |
-
'reason': "Directly matched your preferred breed"
|
323 |
-
})
|
324 |
-
|
325 |
-
# 添加相似品種
|
326 |
-
for breed_name, similarity in similar_breeds:
|
327 |
-
if breed_name != preferred_breed:
|
328 |
-
health_score = self._calculate_health_score(breed_name)
|
329 |
-
noise_info = breed_noise_info.get(breed_name, {
|
330 |
-
"noise_level": "Unknown",
|
331 |
-
"noise_notes": "No noise information available"
|
332 |
-
})
|
333 |
-
|
334 |
-
# 調整相似品種分數計算
|
335 |
-
base_similarity = similarity * 0.6
|
336 |
-
health_factor = health_score * 0.2
|
337 |
-
noise_factor = self._calculate_noise_similarity(preferred_breed, breed_name) * 0.2
|
338 |
-
|
339 |
-
# 確保相似品種分數不會超過偏好品種
|
340 |
-
final_score = min(0.95, base_similarity + health_factor + noise_factor)
|
341 |
-
|
342 |
-
matches.append({
|
343 |
-
'breed': breed_name,
|
344 |
-
'score': final_score,
|
345 |
-
'is_preferred': False,
|
346 |
-
'similarity': similarity,
|
347 |
-
'health_score': health_score,
|
348 |
-
'noise_level': noise_info['noise_level'],
|
349 |
-
'reason': f"Similar to {preferred_breed} in characteristics, health profile, and noise level"
|
350 |
-
})
|
351 |
-
else:
|
352 |
-
matches = self._general_matching(description, top_n)
|
353 |
-
|
354 |
-
return sorted(matches,
|
355 |
-
key=lambda x: (-int(x.get('is_preferred', False)),
|
356 |
-
-x['score'], # 降序排列
|
357 |
-
x['breed']))[:top_n]
|
358 |
-
|
359 |
-
def _detect_breed_preference(self, description: str) -> Optional[str]:
|
360 |
-
"""檢測用戶是否提到特定��種"""
|
361 |
-
description_lower = description.lower()
|
362 |
-
|
363 |
-
for breed_info in self.dog_data:
|
364 |
-
breed_name = breed_info[1]
|
365 |
-
normalized_breed = breed_name.lower().replace('_', ' ')
|
366 |
-
|
367 |
-
if any(phrase in description_lower for phrase in [
|
368 |
-
f"love {normalized_breed}",
|
369 |
-
f"like {normalized_breed}",
|
370 |
-
f"prefer {normalized_breed}",
|
371 |
-
f"want {normalized_breed}",
|
372 |
-
normalized_breed
|
373 |
-
]):
|
374 |
-
return breed_name
|
375 |
-
|
376 |
-
return None
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|