KG Construction Using LLMs
summary
An open‑source pipeline called iText2KG incrementally turns documents into knowledge graphs by distilling text into semantic blocks, extracting entities and relations with embeddings + cosine similarity (merge threshold ≈ 0.7), and pushing results to Neo4j for visualization and querying. A parallel variant (ATOM) models observation vs validity times, decomposes input into atomic fact decomposition chunks (~400 tokens), runs 8 threads with batch size 40, and reports ~18% higher exhaustivity, ~17% better stability and over 90% latency reduction.
AI-written summary · about
chapters (13)
- 0:00Introduction and context
- 1:15Overview of ITX2KG method
- 3:03ITX2KG core algorithm
- 7:04Advantages and drawbacks of ITX2KG
- 8:56Introduction to ATM method
- 10:46ATM practical example
- 13:39ATM performance results
- 14:28Q&A session
- 17:04Hands-on demo setup
- 19:14Building temporal knowledge graph
- 20:20New 4J integration
- 22:12Insights from knowledge graph
- 23:39Final Q&A and closing
related talks
30mBuilding Knowledge Graphs With LLMs
46mKnowledge Graphs + LLMs: What's the latest in research?
28mThe Present and Future of AI-Enabled Knowledge Graph Generation
14mTalk With Your Documents Via Docs2KG
20mPractical GraphRAG: Making LLMs smarter with Knowledge Graphs
53mPractical GraphRAG - Making LLMs smarter with Knowledge Graphs
31mBeyond Vectors: Evolving GenAI through Transformative Tools and Methods
36mGraphRAG Explained 🔥 Building Smarter LLMs