LLMs often hallucinate or lack access to private data because they are trained to be helpful, not necessarily factual.
The RAG pattern improves LLM accuracy by first retrieving relevant information from a database to provide as context for the answer.
Knowledge graphs model data as a network of entities and relationships, making complex connections intuitive and easy to query.
LLMs can automatically extract structured entities and relationships from unstructured documents to populate a knowledge graph.
An application ingests a conference agenda and uses an LLM to automatically build a knowledge graph of speakers and their talks.
The GraphRAG pattern uses vector search to find entry points into the graph and then traverses relationships to gather richer, more relevant context.
A Jupyter notebook demonstrates how to use LangChain and Neo4j to execute a GraphRAG query that avoids LLM hallucinations.
This approach provides rich context, enables explainability by tracing data sources, and allows for graph enrichment with clustering algorithms.
You can guide the LLM's extraction process with a predefined schema and validate its output against a human-created baseline for accuracy.
24:44Tomaz Bratanic
46:51
21:17Carl Lapierre
26:25Ashish Sharma
29:40Jodie Burchell
58:00Vijay Krishan Gupta & Gauravdeep Singh Lotey
31:50Cedric Clyburn & Legare Kerrison
24:42Martin O'Hanlon
.png?w=240&auto=compress,format)
.png?w=240&auto=compress,format)
Jobs that call for the skills explored in this talk.
Understanding Recruitment Group
Barcelona, Spain
auteega GmbH
Deutsches Medizinrechenzentrum GmbH