Update response_processor.py

response_processor.py

@@ -315,20 +315,80 @@ class ResponseProcessor:
     def _critical_format_preprocess(self, response: str) -> str:
         """
         關鍵格式預處理，處理最常見的格式問題
-
+    
         Args:
             response: 原始回應
-
+    
         Returns:
             str: 預處理後的回應
         """
         if not response:
             return response
-
+    
         try:
             import re
+            # 移除各種形式的 confirmed
+            confirmed_patterns = [
+                r'\bconfirmed\s+',  # "confirmed cars" -> "cars"
+                r'\b(\d+)\s+confirmed\s+([a-zA-Z\s]+)',  # "12 confirmed cars" -> "12 cars"
+                r'\b(one|two|three|four|five|six|seven|eight|nine|ten|eleven|twelve)\s+confirmed\s+([a-zA-Z\s]+)',  # "twelve confirmed cars" -> "twelve cars"
+            ]
+            
+            for pattern in confirmed_patterns:
+                if pattern == r'\bconfirmed\s+':
+                    response = re.sub(pattern, '', response, flags=re.IGNORECASE)
+                else:
+                    response = re.sub(pattern, r'\1 \2', response, flags=re.IGNORECASE)
+    
+            # 數字轉文字的完整字典
+            number_conversions = {
+                '0': 'zero', '1': 'one', '2': 'two', '3': 'three', '4': 'four', '5': 'five',
+                '6': 'six', '7': 'seven', '8': 'eight', '9': 'nine', '10': 'ten',
+                '11': 'eleven', '12': 'twelve', '13': 'thirteen', '14': 'fourteen', '15': 'fifteen',
+                '16': 'sixteen', '17': 'seventeen', '18': 'eighteen', '19': 'nineteen', '20': 'twenty'
+            }
+    
+            # 強化數字替換邏輯 - 處理各種語法結構
+            for digit, word in number_conversions.items():
+                # 模式1: 數字 + 名詞 (如 "3 cars", "12 people")
+                pattern1 = rf'\b{digit}\s+([a-zA-Z]+(?:\s+[a-zA-Z]+)*)\b'
+                response = re.sub(pattern1, rf'{word} \1', response)
+    
+                # 模式2: 數字 + visible/present + 名詞 (如 "3 visible traffic lights")
+                pattern2 = rf'\b{digit}\s+(visible|present|apparent|evident)\s+([a-zA-Z]+(?:\s+[a-zA-Z]+)*)\b'
+                response = re.sub(pattern2, rf'{word} \1 \2', response, flags=re.IGNORECASE)
+    
+                # 模式3: 介詞 + 數字 + 名詞 (如 "against a backdrop of 3 visible traffic lights")
+                pattern3 = rf'\b(against|with|featuring|including|containing)\s+(?:a\s+backdrop\s+of\s+)?{digit}\s+([a-zA-Z]+(?:\s+[a-zA-Z]+)*)\b'
+                response = re.sub(pattern3, rf'\1 {word} \2', response, flags=re.IGNORECASE)
+    
+                # 模式4: 複合描述中的數字 (如 "featuring twelve confirmed cars and 3 confirmed persons")
+                pattern4 = rf'\b(and|,)\s+{digit}\s+([a-zA-Z]+(?:\s+[a-zA-Z]+)*)\b'
+                response = re.sub(pattern4, rf'\1 {word} \2', response, flags=re.IGNORECASE)
+                
+            grammar_fixes = [
+                # persons -> people 的全面修正
+                (r'\b(\d+|one|two|three|four|five|six|seven|eight|nine|ten|eleven|twelve|thirteen|fourteen|fifteen|sixteen|seventeen|eighteen|nineteen|twenty)\s+persons\b', r'\1 people'),
+                (r'\bmultiple\s+persons\b', 'multiple people'),
+                (r'\bseveral\s+persons\b', 'several people'),
+                (r'\bmany\s+persons\b', 'many people'),
+                (r'\ba\s+few\s+persons\b', 'a few people'),
+                (r'\bsome\s+persons\b', 'some people'),
+                (r'\bvarious\s+persons\b', 'various people'),
+                (r'\bnumerous\s+persons\b', 'numerous people'),
+                
+                # 修正語法結構問題
+                (r'\bvisible\s+traffic\s+lights\b', 'traffic lights visible'),
+                (r'\bpresent\s+traffic\s+lights\b', 'traffic lights present'),
+                (r'\bapparent\s+traffic\s+lights\b', 'traffic lights apparent'),
+                
+                # 修正重複的形容詞結構
+                (r'\b(visible|present|apparent|evident)\s+(visible|present|apparent|evident)\s+', r'\1 '),
+            ]
+    
+            for pattern, replacement in grammar_fixes:
+                response = re.sub(pattern, replacement, response, flags=re.IGNORECASE)
 
-            # 第一優先級：處理斜線問題
             # 首先處理已知的斜線組合，使用形容詞替換
             for slash_combo, replacement in self.slash_replacements.items():
                 if slash_combo.lower() in response.lower():
@@ -339,13 +399,11 @@ class ResponseProcessor:
                         replacement_formatted = replacement.title()
                     else:
                         replacement_formatted = replacement
-
+    
                     # 執行替換（不區分大小寫）
                     response = re.sub(re.escape(slash_combo), replacement_formatted, response, flags=re.IGNORECASE)
-                    self.logger.debug(f"Replaced slash pattern '{slash_combo}' with '{replacement_formatted}'")
-
+    
             # 處理其他未預定義的斜線模式
-            # 標準斜線模式：word/word
             slash_pattern = r'\b([a-zA-Z]+)/([a-zA-Z]+)\b'
             matches = list(re.finditer(slash_pattern, response))
             for match in reversed(matches):  # 從後往前處理避免位置偏移
@@ -356,22 +414,19 @@ class ResponseProcessor:
                 else:
                     replacement = word2
                 response = response[:match.start()] + replacement + response[match.end():]
-                self.logger.debug(f"Replaced general slash pattern '{match.group(0)}' with '{replacement}'")
-
-            # 第二優先級：處理底線格式
+    
             # 首先處理已知的底線組合
             for underscore_combo, replacement in self.underscore_replacements.items():
                 if underscore_combo in response:
                     response = response.replace(underscore_combo, replacement)
-                    self.logger.debug(f"Replaced underscore pattern '{underscore_combo}' with '{replacement}'")
-
+    
             # 處理三個詞的底線組合：word_word_word → word word word
             response = re.sub(r'\b([a-z]+)_([a-z]+)_([a-z]+)\b', r'\1 \2 \3', response)
-
+    
             # 處理任何剩餘的底線模式：word_word → word word
             response = re.sub(r'\b([a-zA-Z]+)_([a-zA-Z]+)\b', r'\1 \2', response)
-
-            # 第三優先級：修正不完整句子
+    
+            # 確保句子的完整性
             incomplete_sentence_fixes = [
                 (r'\bIn\s*,\s*', 'Throughout the area, '),
                 (r'\bOverall,\s+exudes\b', 'Overall, the scene exudes'),
@@ -379,35 +434,15 @@ class ResponseProcessor:
                 (r'\bwith its lights turned illuminating\b', 'with its lights illuminating'),
                 (r'\bwhere it stands as\b', 'where it stands as'),
             ]
-
+    
             for pattern, replacement in incomplete_sentence_fixes:
                 response = re.sub(pattern, replacement, response, flags=re.IGNORECASE)
-
-            # 第四優先級：語法修正處理(像是person and people)
-            grammar_fixes = [
-                (r'\b(\d+)\s+persons\b', r'\1 people'),
-                (r'\bone\s+persons\b', 'one person'),
-                (r'\btwo\s+persons\b', 'two people'),
-                (r'\bthree\s+persons\b', 'three people'),
-                (r'\bfour\s+persons\b', 'four people'),
-                (r'\bfive\s+persons\b', 'five people'),
-                (r'\bsix\s+persons\b', 'six people'),
-                (r'\bseven\s+persons\b', 'seven people'),
-                (r'\beight\s+persons\b', 'eight people'),
-                (r'\bnine\s+persons\b', 'nine people'),
-                (r'\bten\s+persons\b', 'ten people'),
-                (r'\bmultiple\s+persons\b', 'multiple people'),
-                (r'\bseveral\s+persons\b', 'several people'),
-                (r'\bmany\s+persons\b', 'many people'),
-                (r'\ba\s+few\s+persons\b', 'a few people'),
-                (r'\bsome\s+persons\b', 'some people')
-            ]
-
-            for pattern, replacement in grammar_fixes:
-                response = re.sub(pattern, replacement, response, flags=re.IGNORECASE)
-
+    
+            # 清理多餘空格並確保格式一致性 
+            response = re.sub(r'\s+', ' ', response).strip()
+    
             return response
-
+    
         except Exception as e:
             self.logger.warning(f"Error in critical format preprocessing: {str(e)}")
             return response
@@ -1158,4 +1193,4 @@ class ResponseProcessor:
             "suffixes_to_remove_count": len(self.suffixes_to_remove),
             "repetitive_patterns_count": len(self.repetitive_patterns),
             "initialization_status": "success"
-        }
+        }