Leveraging Knowledge Graphs for RAG: A Smarter Approach to Contextual AI Applications
summary
Concrete claim: knowledge graphs cut hallucinations by scoping and returning the exact evidence used to answer queries. Demo uses Neo4j with named entity recognition over BBC articles, plus a local Llama 33 instance that consumes the retrieved context and generates answers. The pipeline finds two documents for “Ernie Weiss” and returns a precise, source-linked response. Comparative points: graphs trade ingestion and ontology upkeep for reduced hallucinations and easier explainability versus GPU‑accelerated vector stores.
AI-written summary · about
chapters (6)
- 0:00Knowledge graphs and RAG intro
- 1:02Vector vs graph databases
- 11:15Token prediction vs relationships
- 20:10Explainable AI with graphs
- 27:56Non-AI graph uses
- 32:35Key takeaways & Q&A
related talks
36mAgentic GraphRAG: AI’s Logical Edge
27mLies, Damn Lies, and AIs
39mEnhancing RAG with Neo4j Knowledge Graph
42mMaking OSS AI Real with Knowledge Graphs
28mEnhancing LLMs with Graph Technology
53mPractical GraphRAG - Making LLMs smarter with Knowledge Graphs
21mAgentic GraphRAG: Autonomous Knowledge Graph Construction and Adaptive Retrieval
27mAdaptive Knowledge Graphs for Dynamic RAG Systems