Diffbot reposted this
Graph RAG or vector-based RAG? There's some discussions (debate) around which is better. (but we're not continuing this conversation...) Thanks to Tomaz Bratanic, one of the thought leaders and trailblazers in the graph world with cool applications, we jointly explore a project that realizes and visualizes the idea of"bringing unstructured and structured data" together. You'll see side-by-side comparisons of AI-generated responses between vector-only searches and vector+knowledge graph in the video, showing how the latter can outperform the former. Neo4j can store both graph data and text embeddings for vector-based searches, and Diffbot provides verified data from the web — the exact opposite of hallucinated LLM-based sources, so AI applications don't suffer from unreliable outputs. To see the cool network visualization in the app, jump to 3:51 and check out Tomaz and Anej's work. You can always go back and see how we use Diffbot's Article, Natural Language Processing, and Enhance APIs to enrich knowledge graphs with accurate and reliable information. p.s. Who's also going to AI Engineer World's Fair the next 2 days in SF? DM if you're around! #neo4j #diffbot #graphrag