New course with Neo4J: Knowledge Graphs for RAG

I'm delighted to introduce knowledge drafts for RAG, built in partnership with Neo4J and taught by Andreas Kollegar. Knowledge drafts are an important technology for many large businesses, including large web search engines and e commerce sites, but the technology seems highly underrated in academic AI. Is technology worth learning about? And in this course you learn about knowledge drafts in the context of applying it with large language models. Knowledge graphs give a way to store and organize data that emphasizes the relationship between things. For example, a knowledge graph storing information about movies might have actors as one type of node in the graph and movies as another type of node and edges between the actors and movies they appeared in. So the edges indicate the relationships between actors and movies. Note the word graph here is in the computer science data structure sense of the word, with nodes often drawn as little circles and relationships often drawn as edges. This is as opposed to the high school sense of the word graph referring to a chart with say, x and y axes. So if you're building a RAG or retrieval augmented generation system that you want to answer questions about movies and actors, then you can use this data structure to quickly pull up relevant information as context for your LLM. Another example you see in this course will be building a knowledge graph for finance, where the nodes and edges show what investment firms invested in what other companies Knowledge drafts have been a key part of how web search engines pull up relevant information. Many of the cards you see on the side when you do a Google or Bing search are from knowledge drafts, so it's exciting to see retrieval using knowledge drafts now being adapted to RAG with OLMS as well. I'm delighted to be joined by Andreas Holliger, developer evangelist for generative AI and Neo4j. Neo4j is probably the leading graph databases company, and it had helped define the entire category. And Arges was actually employee number 10 in the company, which has since grown to almost 1,000 people. His recent focus has been integrating generative AI tools into knowledge graphs, and so I think you may well be the ideal person to help you build intuition about this important but underappreciated tool in AI.

Thanks Andrew, it's great to be here today, and I'm really looking forward to showing you how knowledge graphs can improve your generative AI applications by providing more context for chat. In this course you'll learn how to build and query a knowledge graph. You'll start by learning how to query a fun database of movie information using Cypher Neo4j's query language. If you've ever used SQL, some aspects of Cypher will be very familiar to you. You'll then go on to use Cypher LangChain and an embeddings model to build a knowledge graph of some financial documents that companies file with the sec, the securities and Exchange Commission. That's a US Government agency that regulates and oversees markets and protects investors. Once you've built this database, you'll use it in combination with LangChain to ask questions of the SEC data like what does this company do? And who are its major investors? Analyzing publicly available documents like this is actually a very common use case of knowledge graphs, and adding an LLM to the mix can really help with that research.

So I hope you enjoy the course and come away able to use knowledge graphs to build something really noteworthy.