
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引擎和大语言模型的逻辑推理与问答框架,专为构建专业领域知识库中的逻辑推理和事实问答解决方案而设计。它通过知识与块互索引结构、概念语义推理的知识对齐、模式约束的知识构建以及逻辑形式引导的混合推理和检索等核心功能,有效克服了传统RAG向量相似度计算模型的模糊性和由OpenIE引入的噪音问题。特别适用于需要处理复杂逻辑推理和多跳事实查询的专业场景,如医疗、法律或金融等领域,能够显著提升专业知识服务的质量和准确性。
Python
Apache License 2.08.8k
Stars
684
Forks
73
Watchers
163
Issues
Star 增长
今日0
近 7 天+14
近 30 天+80
综合评分69.51
默认分支master