From 2ac66c3531f95c380816a32f845a23e864fb6c43 Mon Sep 17 00:00:00 2001
From: yangdx <yangdx@znipower.com>
Date: Sat, 12 Apr 2025 20:45:41 +0800
Subject: [PATCH] Remove chinese quotes in entity name

---
 lightrag/operate.py | 18 +++++++++---------
 lightrag/utils.py   |  9 ++++++++-
 2 files changed, 17 insertions(+), 10 deletions(-)

diff --git a/lightrag/operate.py b/lightrag/operate.py
index ceec7b61..8c6688aa 100644
--- a/lightrag/operate.py
+++ b/lightrag/operate.py
@@ -165,7 +165,7 @@ async def _handle_single_entity_extraction(
         return None
 
     # Normalize entity name
-    entity_name = normalize_extracted_info(entity_name)
+    entity_name = normalize_extracted_info(entity_name, is_entity=True)
 
     # Clean and validate entity type
     entity_type = clean_str(record_attributes[2]).strip('"')
@@ -176,7 +176,7 @@ async def _handle_single_entity_extraction(
         return None
 
     # Clean and validate description
-    entity_description = clean_str(record_attributes[3]).strip('"')
+    entity_description = clean_str(record_attributes[3])
     entity_description = normalize_extracted_info(entity_description)
 
     if not entity_description.strip():
@@ -202,20 +202,20 @@ async def _handle_single_relationship_extraction(
     if len(record_attributes) < 5 or record_attributes[0] != '"relationship"':
         return None
     # add this record as edge
-    source = clean_str(record_attributes[1]).strip('"')
-    target = clean_str(record_attributes[2]).strip('"')
+    source = clean_str(record_attributes[1])
+    target = clean_str(record_attributes[2])
 
     # Normalize source and target entity names
-    source = normalize_extracted_info(source)
-    target = normalize_extracted_info(target)
+    source = normalize_extracted_info(source, is_entity=True)
+    target = normalize_extracted_info(target, is_entity=True)
 
-    edge_description = clean_str(record_attributes[3]).strip('"')
+    edge_description = clean_str(record_attributes[3])
     edge_description = normalize_extracted_info(edge_description)
 
-    edge_keywords = clean_str(record_attributes[4]).strip('"')
+    edge_keywords = clean_str(record_attributes[4]).strip('"').strip("'")
     edge_source_id = chunk_key
     weight = (
-        float(record_attributes[-1].strip('"'))
+        float(record_attributes[-1].strip('"').strip("'"))
         if is_float_regex(record_attributes[-1])
         else 1.0
     )
diff --git a/lightrag/utils.py b/lightrag/utils.py
index 6b9b07fa..8473ea81 100644
--- a/lightrag/utils.py
+++ b/lightrag/utils.py
@@ -1006,7 +1006,7 @@ def get_content_summary(content: str, max_length: int = 250) -> str:
     return content[:max_length] + "..."
 
 
-def normalize_extracted_info(name: str) -> str:
+def normalize_extracted_info(name: str, is_entity = False) -> str:
     """Normalize entity/relation names and description with the following rules:
     1. Remove spaces between Chinese characters
     2. Remove spaces between Chinese characters and English letters/numbers
@@ -1040,6 +1040,13 @@ def normalize_extracted_info(name: str) -> str:
     # Remove English quotation marks from the beginning and end
     name = name.strip('"').strip("'")
 
+    if is_entity:
+        # remove Chinese quotes
+        name = name.replace("“", "").replace("”", "").replace("‘", "").replace("’", "")
+        # remove English queotes in and around chinese
+        name = re.sub(r"['\"]+(?=[\u4e00-\u9fa5])", "", name)
+        name = re.sub(r"(?<=[\u4e00-\u9fa5])['\"]+", "", name)
+
     return name