Graph-Driven Knowledge Retrieval: Neo4j for Healthcare Building Codes
summary
Neo4j is used as the vector-backed knowledge graph in a RAG pipeline with OpenAI and LangSmith evaluators to test retrieval+generation on healthcare building codes. Three indexing approaches are compared: simple text-splitter chunks, a Raptor three-layer summary clustering, and an LLM graph Transformer that extracts entities and links. Ground-truth tests (consultation rooms = 13) show text-splitter and Raptor score near zero on faithfulness and correctness; the graph Transformer predicts 10 (closer but still incorrect), so nothing is production-ready yet.
AI-written summary · about
chapters (7)
- 0:00Introduction to healthcare codes
- 1:32Indexing strategies for Neo4j
- 3:38What is LSmith and its workflow
- 6:31Evaluation workflows in LSmith
- 9:03Evaluator types and prompts
- 10:51Experiment results and analysis
- 14:21Summary and next steps
related talks
27mAdaptive Knowledge Graphs for Dynamic RAG Systems
21mAgentic GraphRAG: Autonomous Knowledge Graph Construction and Adaptive Retrieval
12mAgentic GraphRAG: Autonomous Knowledge Graph Construction and Adaptive Retrieval
33mSmarter Industrial Fault Diagnosis with Knowledge Graphs and Large Language Models
16mEnhancing Retrieval-Augmented Generation with GraphRAG Patterns in Neo4j
01h 17mGraph Database and Retrieval Augmented Generation
42mMaking OSS AI Real with Knowledge Graphs
36mAgentic GraphRAG: AI’s Logical Edge