feat: 将漏掉的身份证号和社会安全号补上

backend/app/core/document_handlers/ner_processor.py

@@ -7,7 +7,7 @@ from ...core.config import settings
 from ..utils.json_extractor import LLMJsonExtractor
 from ..utils.llm_validator import LLMResponseValidator
 import re
-from .regs.entity_regex import extract_id_number_entities, extract_social_credit_code_entities
+from .regs.entity_regex import extract_id_number_entities, extract_social_credit_code_entities, extract_case_number_entities
 from .extractors.ner_extractor import NERExtractor
 from pypinyin import pinyin, Style
 
@@ -722,7 +722,8 @@ class NerProcessor:
 
         regex_entity_extractors = [
             extract_id_number_entities,
-            extract_social_credit_code_entities
+            extract_social_credit_code_entities,
+            extract_case_number_entities
         ]
         for extractor in regex_entity_extractors:
             mapping = extractor(chunk)
@@ -733,6 +734,38 @@ class NerProcessor:
 
         return mapping_pipeline
     
+    def build_mapping_regex_only(self, chunk: str) -> list[Dict[str, str]]:
+        """
+        Build mapping using only regex-based extraction (no NER, no LLM)
+        
+        Args:
+            chunk: Text chunk to process
+            
+        Returns:
+            List of entity mappings
+        """
+        mapping_pipeline = []
+        
+        # Use regex-based extraction for IDs, codes, and case numbers
+        regex_entity_extractors = [
+            extract_id_number_entities,
+            extract_social_credit_code_entities,
+            extract_case_number_entities
+        ]
+        
+        for extractor in regex_entity_extractors:
+            mapping = extractor(chunk)
+            if mapping and LLMResponseValidator.validate_regex_entity(mapping):
+                mapping_pipeline.append(mapping)
+                logger.info(f"Regex extraction: Added mapping from {extractor.__name__}")
+            elif mapping:
+                logger.warning(f"Invalid regex entity mapping format: {mapping}")
+            else:
+                logger.debug(f"No entities found by {extractor.__name__}")
+
+        logger.info(f"Regex-only extraction: Found {len(mapping_pipeline)} mappings")
+        return mapping_pipeline
+    
     def build_mapping_llm_only(self, chunk: str) -> list[Dict[str, str]]:
         """
         Build mapping using only LLM (no NER)
@@ -758,10 +791,11 @@ class NerProcessor:
             if mapping:
                 mapping_pipeline.append(mapping)
 
-        # Include regex-based extraction for IDs and codes
+        # Include regex-based extraction for IDs, codes, and case numbers
         regex_entity_extractors = [
             extract_id_number_entities,
-            extract_social_credit_code_entities
+            extract_social_credit_code_entities,
+            extract_case_number_entities
         ]
         for extractor in regex_entity_extractors:
             mapping = extractor(chunk)
@@ -792,10 +826,11 @@ class NerProcessor:
             mapping_pipeline.append(ner_mapping)
             logger.info(f"NER-only extraction: Added {len(ner_entities)} entities")
         
-        # Still include regex-based extraction for IDs and codes
+        # Still include regex-based extraction for IDs, codes, and case numbers
         regex_entity_extractors = [
             extract_id_number_entities,
-            extract_social_credit_code_entities
+            extract_social_credit_code_entities,
+            extract_case_number_entities
         ]
         for extractor in regex_entity_extractors:
             mapping = extractor(chunk)
@@ -1179,7 +1214,12 @@ class NerProcessor:
             chunk_mappings.append(ner_mapping)
             logger.info(f"Added {len(ner_entities)} NER entities to mappings")
         
-        logger.info(f"Final chunk mappings: {chunk_mappings}")
+        logger.info(f"NER-only mappings: {chunk_mappings}")
+
+        regex_mapping = self.build_mapping_regex_only(merged_text)
+        logger.info(f"Regex mapping: {regex_mapping}")
+        chunk_mappings.extend(regex_mapping)
+
             
         unique_entities = self._merge_entity_mappings(chunk_mappings)
         logger.info(f"Unique entities: {unique_entities}")

backend/app/core/document_handlers/regs/entity_regex.py

@@ -15,4 +15,13 @@ def extract_social_credit_code_entities(chunk: str) -> dict:
     entities = []
     for match in re.findall(credit_pattern, chunk):
         entities.append({"text": match, "type": "统一社会信用代码"})
+    return {"entities": entities} if entities else {}
+
+def extract_case_number_entities(chunk: str) -> dict:
+    """Extract case numbers and return in entity mapping format."""
+    # Pattern for Chinese case numbers: (2022)京 03 民终 3852 号, （2020）京0105 民初69754 号
+    case_pattern = r'[（(]\d{4}[）)][^\d]*\d+[^\d]*\d+[^\d]*号'
+    entities = []
+    for match in re.findall(case_pattern, chunk):
+        entities.append({"text": match, "type": "案号"})
     return {"entities": entities} if entities else {}