# honeyandme/RAGQnASystem

本项目设计并实现了一个基于知识图谱 RAG 与大语言模型的医疗智能问答系统。系统以 DiseaseKG 医疗数据集为基础，依托 Neo4j 构建包含 4.4 万实体与 31 万关系的领域知识图谱；结合 BERT 命名实体识别与 32B 大模型意图识别，通过精确的图谱检索与受控的答案生成，有效缓解大模型在医疗场景中的幻觉问题，显著提升回答的准确性与可靠性。

Repository: https://github.com/honeyandme/RAGQnASystem
Canonical: https://ross.abutalabs.com/products/ragqnasystem
Language: Jupyter Notebook
License Family: other
Last push: 2026-05-29T05:18:37+00:00

## Health v2 (maintenance only)
Score: 62/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 84, release rhythm 35, longevity 59
- inputs: {"age_days": 839, "days_push": 96, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1415, forks 149 (observed 2026-08-28T04:04:39.798237+00:00)

## What it is
A medical intelligent question-answering system combining knowledge-graph RAG with large language models, built on the DiseaseKG dataset with a Neo4j graph of 44k entities and 310k relations. It uses BERT+RNN for named entity recognition, a local LLM via ollama for intent recognition and streaming answer generation, and a Streamlit UI, aiming to reduce hallucinations in medical Q&A.

## Use cases
- build a medical question answering chatbot
- answer health questions from a disease knowledge graph
- reduce LLM hallucinations with structured knowledge retrieval
- run a local RAG system with ollama
- do named entity recognition on medical questions
- query a Neo4j medical knowledge graph with Cypher
- classify user question intents with an LLM
- learn how to combine KG retrieval with LLM generation

## When to choose
- you need accurate, grounded answers for medical/disease questions in Chinese
- you want a self-hosted RAG pipeline using a knowledge graph instead of vector search
- you want a complete reference implementation with NER, intent routing, and a web UI
- you have Neo4j and a local LLM available and want to study KG-RAG architecture

## When to avoid
- you need a production-grade, clinically validated medical system
- you need English-language medical Q&A rather than Chinese
- you want a simple vector-database RAG setup
- you need a hosted SaaS solution or a permissive license (none is provided)

## Facets
- artifact type: application
- maturity: active
- function: rag, nlp, chatbot, machine-learning, search-engine, llm-inference
- domain: healthcare, large-language-models, chatbots, artificial-intelligence
- platform: python, self-hosted
- tags: knowledge-graph, neo4j, bert-ner, medical-qa, ollama, streamlit, intent-recognition, hallucination-mitigation, natural-language-processing, retrieval-augmented-generation, docker

## Member repositories
- honeyandme/RAGQnASystem (main) score 62

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:39.798237+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:38:09.021730+00:00, confidence not recorded.
  - readme: https://github.com/honeyandme/RAGQnASystem (fetched 2026-08-28T04:04:39.798237+00:00, sha 31190d764ed9)
- Data as of 2026-08-30T08:39:29.467469+00:00.