Sign up to save tools and stay up to date with the latest in AI
bg
bg
1

GraphRAG: Using Knowledge in Unstructured Data to Build Apps with LLMs - Graphlit

May 21, 2024 - graphlit.com
The article discusses the integration of Language Learning Models (LLMs) with unstructured data through Retrieval Augmented Generation (RAG) for various business use cases. It explains how businesses can use knowledge graphs to model relationships between people, organizations, places, and events, and how LLMs can extract these entities and their inter-relationships. The article also introduces GraphRAG, a pattern that extends the standard RAG pipeline, which can enrich the retrieval stage of the pipeline and provide greater context about the referenced entities in the LLM response.

The author provides a detailed walkthrough of the code required to build a demo application using Graphlit, a platform that allows users to build a knowledge graph automatically using OpenAI GPT-4o for entity extraction. The demo application can be used to ingest files from SharePoint, visualize a knowledge graph, and create a GraphRAG conversation. The article concludes by inviting readers to share their requirements for knowledge graphs, LLMs, and GraphRAG.

Key takeaways:

  • Graphlit's GraphRAG is a pattern that extends the standard Retrieval Augmented Generation (RAG) pipeline, allowing businesses to extract and model relationships between entities from unstructured data sources like Microsoft SharePoint, email, Slack, Microsoft Teams, and Zoom recordings.
  • GraphRAG can be used for filtering and providing context in the RAG pipeline, enhancing the quality of responses from Language Learning Models (LLMs) by providing additional information about the entities that the user is interested in.
  • Graphlit provides a demo application on GitHub that demonstrates how to build a knowledge graph using OpenAI GPT-4o for entity extraction and how to create a GraphRAG conversation.
  • Graphlit's platform allows for the creation of recurrent feeds that check for new files in specified SharePoint folders every minute, and the extracted entities can be enriched with additional information from external APIs like Wikipedia.
View Full Article

Comments (0)

Be the first to comment!