From acde3bd32f07bf02839a21e8fe5b4e69bfca2251 Mon Sep 17 00:00:00 2001
From: lyg <1543117173@qq.com>
Date: 星期三, 14 五月 2025 10:37:00 +0800
Subject: [PATCH] docx文档拆分，文档段落实体词提取，存入mysql数据库。
---
knowledgebase/doc/doc_processor.py | 133 +++++++++++++++++++++++++++++---------------
1 files changed, 87 insertions(+), 46 deletions(-)
diff --git a/knowledgebase/doc/doc_processor.py b/knowledgebase/doc/doc_processor.py
index 7dccb8b..bc0e0dd 100644
--- a/knowledgebase/doc/doc_processor.py
+++ b/knowledgebase/doc/doc_processor.py
@@ -1,65 +1,106 @@
# -*- coding: utf-8 -*-
# @file: doc_processor.py
# @author: lyg
-# @date: 20250427
+# @date: 2025-5-13
# @version:
-# @description: 澶勭悊鏂囨。锛屾彁鍙栫珷鑺備俊鎭紝鎻愬彇椤电爜淇℃伅锛屾彁鍙栧疄浣撹瘝锛屽啓鍏ュ浘鏁版嵁搴擄紙neo4j锛夈��
-from knowledgebase.db.neo4j import Neo4jHelper
-from knowledgebase.doc.doc_split import DocSplit
-from knowledgebase.doc.entity_recognition import EntityRecognition
+# @description: 澶勭悊鏂囨。锛屾媶鍒嗘枃妗ｏ紝灏嗘媶鍒嗗悗鐨勭珷鑺備繚瀛樺埌鏁版嵁搴撲腑銆�
+from langchain_core.messages import HumanMessage
+
+from knowledgebase.doc.docx_split import DocSplit
import asyncio
+from knowledgebase.db.doc_db_helper import doc_dbh
+from knowledgebase.doc.entity_helper import entity_helper
+from knowledgebase.doc.entity_recognition import EntityRecognition
+import os.path
+
+from knowledgebase.doc.models import DocInfo, ParagraphInfo
+from knowledgebase.llm import llm
+from knowledgebase.log import Log
+from knowledgebase import utils
class DocProcessor:
- def __init__(self, pdf_file):
- self.doc_split = DocSplit(pdf_file)
- self.entity_recognition = EntityRecognition()
- self.neo4j = Neo4jHelper()
+ def __init__(self, docx_file: str):
+ """
+ 鏂囨。澶勭悊
+ :param docx_file: 瑕佸鐞嗙殑鏂囨。
+ """
+ Log.info(f'寮�濮嬪鐞嗘枃妗ｏ細{docx_file}')
+ self.docx_file = docx_file
+ self.doc_split = DocSplit(docx_file)
+ self.doc_type = self.get_doc_type()
+ self.entity_recognition = EntityRecognition(self.doc_type)
+ self.doc_id = 0
- async def gen_page_entities(self, page_info):
- # 鑾峰彇椤甸潰瀹炰綋璇�
- page_entities = await asyncio.to_thread(lambda: self.entity_recognition.run(page_info.text))
- page_info.entities = page_entities
+ def get_doc_type(self):
+ Log.info(f'璇嗗埆鏂囨。绫诲瀷锛歿self.docx_file}')
+ rules = '锛沑n'.join([f'- {it}锛歿entity_helper.doc_prompt_map[it]}' for it in entity_helper.doc_prompt_map.keys()])
+ msg = HumanMessage(f'''
+# 鎸囦护
+璇蜂粠涓嬮潰鐨勬枃浠跺悕涓瘑鍒枃妗ｇ被鍨嬶紝濡傛灉璇嗗埆澶辫触涓嶈杈撳嚭浠讳綍瀛楃銆�
+鏂囦欢鍚嶏細{os.path.basename(self.docx_file)}
+# 璇嗗埆瑙勫垯
+{rules}
+# 绀轰緥
+閬ユ祴澶х翰
+''')
+ resp = llm.invoke([msg])
+ Log.info(f'璇嗗埆缁撴灉锛歿resp.content}')
+ return resp.content
+
+ async def gen_sect_entities(self, paragraph: ParagraphInfo):
+ # Log.info(f'鐢熸垚绔犺妭瀹炰綋璇嶏細{paragraph.full_text}')
+ # 鑾峰彇绔犺妭瀹炰綋璇�
+ entities = await asyncio.to_thread(lambda: self.entity_recognition.run(paragraph.full_text))
+ Log.info(f'绔犺妭瀹炰綋璇嶏細{entities}')
+ if entities:
+ paragraph.entities = [next(filter(lambda x: x.name == e, entity_helper.entities), None) for e in entities]
+ paragraph.entities = [e for e in paragraph.entities if e]
def process(self):
- # 鍒嗘壒骞跺彂澶勭悊锛屾瘡鎵�10椤�
+ self.doc_split.split()
+ # 鍒嗘壒骞跺彂澶勭悊锛屾瘡鎵�10涓�
batch_size = 10
- for i in range(0, len(self.doc_split.page_infos), batch_size):
- batch_page_infos = self.doc_split.page_infos[i:i + batch_size]
+ for i in range(0, len(self.doc_split.paragraphs), batch_size):
+ batch_paragraphs = self.doc_split.paragraphs[i:i + batch_size]
tasks = []
- for page_info in batch_page_infos:
- tasks.append(self.gen_page_entities(page_info))
- asyncio.run(asyncio.gather(*tasks))
- self.save_to_neo4j()
+ for paragraph in batch_paragraphs:
+ tasks.append(self.gen_sect_entities(paragraph))
- def save_to_neo4j(self):
- """
- 淇濆瓨椤靛拰椤靛疄浣撹瘝鍒皀eo4j鏁版嵁搴撱��
+ async def run():
+ await asyncio.gather(*tasks)
- 1.姣忎竴椤典负涓�涓狽ode锛�
- 2.姣忎竴涓疄浣撹瘝涓轰竴涓狽ode锛�
- 3.椤靛拰瀹炰綋璇嶇洿鎺ュ缓绔嬪叧绯� 椤�->瀹炰綋璇�
- :return:
+ asyncio.run(run())
+ # 淇濆瓨鍒版暟鎹簱
+ self.save_to_db()
+
+ def save_to_db(self):
"""
- for page_info in self.doc_split.page_infos:
- # 鍒涘缓椤佃妭鐐�
- page_node = self.neo4j.create_page_node(page_info)
- entity_nodes = []
- for entity in page_info.entities:
- # 鍒涘缓瀹炰綋璇嶈妭鐐�
- entity_node = self.neo4j.create_entity_node(entity)
- # 寤虹珛鍏崇郴 椤�->瀹炰綋璇�
- self.neo4j.create_page_entity_relationship(page_node, entity_node)
- entity_nodes.append(entity_node)
- if len(entity_nodes) > 0:
- for i in range(len(entity_nodes)):
- prev_entity_node = entity_nodes[i]
- for entity_node in entity_nodes[i + 1:]:
- # 寤虹珛鍏崇郴 涓�椤典腑鐨� 瀹炰綋璇�1->瀹炰綋璇�2
- self.neo4j.create_entity_relationship(prev_entity_node, entity_node)
+ 淇濆瓨娈佃惤鍜屾钀藉疄浣撹瘝鍏崇郴鍒版暟鎹簱銆�
+ """
+ Log.info('淇濆瓨娈佃惤鍜屾钀藉疄浣撹瘝鍏崇郴鍒版暟鎹簱...')
+ with open(self.docx_file, 'rb') as f:
+ file_bytes = f.read()
+ md5 = utils.generate_bytes_md5(file_bytes)
+ doc = DocInfo(os.path.basename(self.docx_file), md5, self.doc_type, self.doc_split.paragraph_tree)
+ self.doc_id = doc_dbh.add_doc(doc)
+ for paragraph in doc.paragraphs:
+ doc_dbh.add_paragraph(self.doc_id, None, paragraph)
+ Log.info('淇濆瓨娈佃惤鍜屾钀藉疄浣撹瘝鍏崇郴鍒版暟鎹簱瀹屾垚')
if __name__ == '__main__':
- pdf_file = "D:/workspace/PythonProjects/KnowledgeBase/doc/XA-5D鏃犱汉鏈烘帰娴嬪ぇ绾诧紙鍏紑锛�111.pdf"
- doc_processor = DocProcessor(pdf_file)
- doc_processor.process()
+ files = [
+ r"D:\workspace\PythonProjects\KnowledgeBase\doc\XA-5D鏃犱汉鏈�1553B鎬荤嚎浼犺緭閫氫俊甯у垎閰嶏紙鍏紑锛�.docx",
+ r"D:\workspace\PythonProjects\KnowledgeBase\doc\XA-5D鏃犱汉鏈哄垎绯荤粺閬ユ祴婧愬寘璁捐鎶ュ憡锛堝叕寮�锛�.docx",
+ r"D:\workspace\PythonProjects\KnowledgeBase\doc\XA-5D鏃犱汉鏈鸿蒋浠剁敤鎴烽渶姹傦紙鍏紑锛�.docx",
+ r"D:\workspace\PythonProjects\KnowledgeBase\doc\XA-5D鏃犱汉鏈洪仴娴嬪ぇ绾诧紙鍏紑锛�.docx",
+ r"D:\workspace\PythonProjects\KnowledgeBase\doc\XA-5D鏃犱汉鏈洪仴娴嬩俊鍙峰垎閰嶈〃锛堝叕寮�锛�.docx",
+ r"D:\workspace\PythonProjects\KnowledgeBase\doc\XA-5D鏃犱汉鏈烘寚浠ゆ牸寮忎笌缂栫爜瀹氫箟锛堝叕寮�锛�.docx",
+ r"D:\workspace\PythonProjects\KnowledgeBase\doc\鎸囦护鏍煎紡(鍏紑).docx"
+ ]
+ for file in files:
+ doc_processor = DocProcessor(file)
+ doc_processor.process()
+
+ # doc_dbh.get_docs()
--
Gitblit v1.9.1