From Unstructured to Connected: Building Knowledge Graphs for GenAI
summary
Shows how the open‑source neo4j-graphrag Python package ingests a 9‑page Wikipedia PDF about climate change, splits text (default 7 chunks, custom section splitter yields 13), uses an LLM schema and GPT-4o for entity+relation extraction, builds both a lexical graph and an entity graph, enforces a schema with graph pruning, writes to Neo4j, and then runs entity resolution (default exact‑name or a spaCy semantic resolver) to merge duplicates (CO2 nodes merged), and allows custom loader/splitter/writer components.
AI-written summary · about
chapters (7)
- 0:00Introduction to neo4j-graphrag
- 2:20Pipeline component breakdown
- 5:05Simple KG pipeline demonstration
- 10:55Schema control and refinement
- 18:52Advanced entity resolution
- 23:30Customizing pipeline components
- 26:44Conclusion and resources
related talks
30mBuilding Knowledge Graphs With LLMs
36mGraphRAG Explained 🔥 Building Smarter LLMs
30mHow to Build the Next Wave of Graph-Based Agentic AI
15mConstructing Knowledge Graphs from Sharepoint Site
30mEntity Resolved Knowledge Graphs
28mBuilding Knowledge Graph-Based Agents with Structured Text Generation
15mGraph-Driven Knowledge Retrieval: Neo4j for Healthcare Building Codes
31mEmpowering Open-Source Generative AI by Integrating the Wikidat...