Session Outline
The session at the Data Innovation Summit 2024 is about how Infomedia utilized the RAG (Retrieval-Augmented Generation) setup for Entity Relation Extraction in Danish. The architecture involves using a Knowledge Graph instead of a vector database for easily fetching relations based on the context of the news articles.
Key Takeaways
- RAG setup and how to replace a vector database with a Knowledge Graph for a custom architecture.
- How to keep the KG updated.
- Prompt engineering and fine-tuning the Large Language Model to incorporate the context from the KG well.
Add comment