Invest in Knowledge Graphs for Improved Accuracy

The podcast discusses the significance of investing in knowledge graphs to improve the accuracy of large language models (LLMs) in answering natural language questions over enterprise SQL databases. The research questions asked were: to what extent can LLMs accurately answer natural language questions over enterprise SQL databases, and to what extent do knowledge graphs improve the accuracy of LLMs? The results showed that using a knowledge graph alongside LLMs increased the accuracy by 3x compared to not using a knowledge graph. With a knowledge graph, easy questions over easy data had a 70% accuracy, while complex questions with more than five tables had a 38% accuracy. The conclusion emphasizes the importance of investing in knowledge graphs and treating business context and semantics as first-class citizens for generative AI applications.