#!/usr/bin/env python3
# -*- coding: utf-8 -*-
import asyncio
from kg_utils import KnowledgeGraphBuilder
async def main():
"""知识图谱构建示例"""
# 您的Silicon Flow API密钥
api_key = "sk-igsxqudwjumptrovmyuuxemhjvhwqxnhegsuuswqpipnxfre"
# 创建知识图谱构建器（使用LLM）
kg_builder_llm = KnowledgeGraphBuilder(api_key=api_key)
# 创建知识图谱构建器（不使用LLM，仅规则）
kg_builder_rule = KnowledgeGraphBuilder()
# 测试文本
test_texts = [
"张三是阿里巴巴公司的CEO，阿里巴巴总部位于杭州。",
"李四在北京大学学习计算机科学，北京大学位于北京市海淀区。",
"王五担任腾讯公司的产品经理，腾讯公司在深圳有分公司。"
]
print("=" * 60)
print("知识图谱构建对比测试")
print("=" * 60)
for i, text in enumerate(test_texts, 1):
print(f"\n【测试文本 {i}】: {text}")
print("-" * 50)
# 使用LLM构建知识图谱
print("🤖 使用LLM构建知识图谱:")
try:
llm_result = await kg_builder_llm.build_graph(text, use_llm=True)
print(f"实体数量: {len(llm_result['entities'])}")
print(f"关系数量: {len(llm_result['relations'])}")
print(f"三元组数量: {len(llm_result['triples'])}")
print("实体:", llm_result['entities'])
print("关系:", llm_result['relations'])
print("三元组:")
for triple in llm_result['triples']:
print(f" - {triple}")
if llm_result['confidence_scores']:
avg_confidence = sum(llm_result['confidence_scores']) / len(llm_result['confidence_scores'])
print(f"平均置信度: {avg_confidence:.3f}")
except Exception as e:
print(f"LLM构建失败: {e}")
print()
# 使用规则构建知识图谱
print("📋 使用规则构建知识图谱:")
try:
rule_result = await kg_builder_rule.build_graph(text, use_llm=False)
print(f"实体数量: {len(rule_result['entities'])}")
print(f"关系数量: {len(rule_result['relations'])}")
print(f"三元组数量: {len(rule_result['triples'])}")
print("实体:", rule_result['entities'])
print("关系:", rule_result['relations'])
print("三元组:")
for triple in rule_result['triples']:
print(f" - {triple}")
if rule_result['confidence_scores']:
avg_confidence = sum(rule_result['confidence_scores']) / len(rule_result['confidence_scores'])
print(f"平均置信度: {avg_confidence:.3f}")
except Exception as e:
print(f"规则构建失败: {e}")
print("=" * 60)
# 获取统计信息
print("\n📊 LLM版本统计信息:")
llm_stats = kg_builder_llm.get_statistics()
for key, value in llm_stats.items():
print(f"{key}: {value}")
print("\n📊 规则版本统计信息:")
rule_stats = kg_builder_rule.get_statistics()
for key, value in rule_stats.items():
print(f"{key}: {value}")
# 导出知识图谱
print("\n💾 导出知识图谱 (JSON格式):")
json_export = kg_builder_llm.export_graph("json")
print(json_export[:500] + "..." if len(json_export) > 500 else json_export)
if __name__ == "__main__":
asyncio.run(main())
