OpenSPG

KAG

OpenSPG

KAG is a logical form-guided reasoning and retrieval framework based on OpenSPG engine and LLMs. It is used to build logical reasoning and factual Q&A solutions for professional domain knowledge bases. It can effectively overcome the shortcomings of the traditional RAG vector similarity calculation model.

AI 简介

KAG是一个基于OpenSPG知识图谱引擎与大语言模型的逻辑形式引导式推理与检索框架,专为专业领域知识库构建高可信的事实问答与多跳逻辑推理能力。其核心采用知识与文本块互索引、概念语义对齐、Schema约束的知识构建、以及逻辑形式驱动的混合检索与推理机制,显著缓解传统RAG的语义歧义和GraphRAG中OpenIE引入的噪声问题。适用于金融、法律、医疗等需强逻辑性与事实可追溯性的垂直领域知识服务场景。

Python
Apache License 2.0
8.9k
Stars
693
Forks
74
Watchers
163
Issues

Star 增长

今日0
近 7 天0
近 30 天+32
综合评分67.72
默认分支master