In this Neo4j Live session, Niels De Jong and Morgan Senechal present the Knowledge Graph Builder App, an innovative application that transforms PDF documents into dynamic knowledge graphs stored in Neo4j. The app leverages OpenAI’s GPT and Diffbot LLM to extract nodes, relationships, and properties from PDF text content, using the Langchain framework to structure this information into a knowledge graph.
The session includes an introduction to Neo4j and LLMs, focusing on how they can be integrated to enhance data visualization and exploration. The key components and architecture of the Knowledge Graph Builder App are discussed, highlighting its ability to upload files from local machines or S3 buckets and select the appropriate LLM model for knowledge graph creation.
The presentation covers the following main sections:
1. **Introduction and Overview**: Niels and Morgan introduce themselves and the motivation behind the Knowledge Graph Builder App. They discuss the growing importance of generative AI and LLMs in creating and consuming knowledge graphs.
2. **Neo4j + LLMs**: An overview of how Neo4j integrates with LLMs, including use cases for graph-powered search, natural language queries, and grounding LLMs in a trusted knowledge graph.
3. **Knowledge Graph Builder**: Detailed explanation of the app’s architecture, which includes data sources, LLMs, Neo4j database, and the ReactJS frontend. The app processes documents by splitting them into chunks, generating embeddings, and extracting entities to create a structured knowledge graph.
4. **Demo**: A live demonstration of the app, showing how to upload PDF documents, YouTube videos, and Wikipedia sources to create a knowledge graph. The demo also features a chatbot interface that allows users to query the knowledge graph using natural language.
5. **Summary and Q&A**: A summary of the session and answers to audience questions about the app’s functionality, scalability, and future enhancements.
The session emphasizes the app’s ease of use, allowing users to create and query knowledge graphs without needing to write code. The app is open-source, and the presenters encourage the community to contribute and extend its capabilities. The source code is available on GitHub, and a public app instance is provided for users to try out.
Overall, the Neo4j Live session provides a comprehensive overview of the Knowledge Graph Builder App, demonstrating its potential to transform unstructured text into valuable, interconnected data that can be easily explored and queried.