使用 LangChain 和 Neo4j 构建知识图谱

在现代 AI 应用中，将文本内容结构化为知识图谱是一种非常实用的方法，可以帮助我们更高效地进行信息查询、知识推理和数据分析。本文将结合 LangChain 官方文档和实战代码，演示如何使用 LangChain + Neo4j 将文本内容自动转换为知识图谱。

环境准备

我们需要以下组件：

Python 3.10+
Neo4j 数据库 (可以通过docker 来安装)
LangChain 及社区扩展模块
统一 AI 模型（如 ChatTongyi 或 Qwen 系列）

首先，安装必要依赖：

复制代码

pip install langchain langchain-neo4j langchain-community

并确保 Neo4j 服务在本地运行：

docker 安装的命令：

复制代码

docker run --name neo4j -p7474:7474 -p7687:7687 -d -e NEO4J_AUTH=neo4j/xxxx  -e NEO4JLABS_PLUGINS='["apoc"]'  -e NEO4J_apoc_export_file_enabled=true -e NEO4J_apoc_import_file_enabled=true  -e NEO4J_apoc_import_file_use__neo4j__config=true neo4j:latest

核心思路

本文示例展示了如何：

使用 LLM（大语言模型）分析文本；
将文本转换为图谱节点和关系；
将生成的图谱写入 Neo4j 数据库。

主要流程图如下：

复制代码

文本 → LLM → 图谱节点/关系 → Neo4j

示例代码解析

下面是完整代码示例：

python 复制代码

import asyncio
import os

from langchain_neo4j import Neo4jGraph
from langchain_community.chat_models import ChatTongyi

os.environ["NEO4J_URI"] = "bolt://localhost:7687"
os.environ["NEO4J_USERNAME"] = "neo4j"
os.environ["NEO4J_PASSWORD"] = "xxxx"

async def test():
    # 初始化 LLM
    llm = ChatTongyi(model="qwen-plus", api_key="YOUR_API_KEY")

    # 连接 Neo4j 图数据库
    graph = Neo4jGraph(refresh_schema=False)

    # 引入图谱转换器
    from langchain_experimental.graph_transformers import LLMGraphTransformer
    llm_transformer = LLMGraphTransformer(llm=llm)

    # 准备文本
    from langchain_core.documents import Document
    text = """
    Marie Curie, born in 1867, was a Polish and naturalised-French physicist and chemist who conducted pioneering research on radioactivity.
    She was the first woman to win a Nobel Prize, the first person to win a Nobel Prize twice, and the only person to win a Nobel Prize in two scientific fields.
    Her husband, Pierre Curie, was a co-winner of her first Nobel Prize, making them the first-ever married couple to win the Nobel Prize and launching the Curie family legacy of five Nobel Prizes.
    She was, in 1906, the first woman to become a professor at the University of Paris.
    """
    documents = [Document(page_content=text)]

    # 将文本转换为图谱文档
    graph_documents = await llm_transformer.aconvert_to_graph_documents(documents)
    print(f"Nodes:{graph_documents[0].nodes}")
    print(f"Relationships:{graph_documents[0].relationships}")

    # 写入 Neo4j
    graph.add_graph_documents(graph_documents)

if __name__ == '__main__':
    asyncio.run(test())

运行结果

python 复制代码

Nodes:[Node(id='Marie Curie', type='Person', properties={}), Node(id='Polish', type='Nationality', properties={}), Node(id='French', type='Nationality', properties={}), Node(id='Physicist', type='Occupation', properties={}), Node(id='Chemist', type='Occupation', properties={}), Node(id='Radioactivity', type='Scientific field', properties={}), Node(id='Nobel Prize', type='Award', properties={}), Node(id='Pierre Curie', type='Person', properties={}), Node(id='University Of Paris', type='Institution', properties={})]
Relationships:[Relationship(source=Node(id='Marie Curie', type='Person', properties={}), target=Node(id='Polish', type='Nationality', properties={}), type='NATIONALITY', properties={}), Relationship(source=Node(id='Marie Curie', type='Person', properties={}), target=Node(id='French', type='Nationality', properties={}), type='NATIONALITY', properties={}), Relationship(source=Node(id='Marie Curie', type='Person', properties={}), target=Node(id='Physicist', type='Occupation', properties={}), type='OCCUPATION', properties={}), Relationship(source=Node(id='Marie Curie', type='Person', properties={}), target=Node(id='Chemist', type='Occupation', properties={}), type='OCCUPATION', properties={}), Relationship(source=Node(id='Marie Curie', type='Person', properties={}), target=Node(id='Radioactivity', type='Scientific field', properties={}), type='RESEARCHED', properties={}), Relationship(source=Node(id='Marie Curie', type='Person', properties={}), target=Node(id='Nobel Prize', type='Award', properties={}), type='AWARDED', properties={}), Relationship(source=Node(id='Marie Curie', type='Person', properties={}), target=Node(id='Nobel Prize', type='Award', properties={}), type='AWARDED_TWICE', properties={}), Relationship(source=Node(id='Marie Curie', type='Person', properties={}), target=Node(id='Pierre Curie', type='Person', properties={}), type='SPOUSE', properties={}), Relationship(source=Node(id='Marie Curie', type='Person', properties={}), target=Node(id='University Of Paris', type='Institution', properties={}), type='PROFESSOR', properties={})]

运行效果

http://localhost:7474/browser/preview/