When Knowledge Graphs Meet LLMs

Knowledge graphs are highly effective for data exploration and the generation ofinsights from complex datasets. However, they can become challenging to manageat scale. In contrast, large language models (LLMs) are highly scalable and proficientin data summarisation, but are susceptible to generating inaccuracies, known ashallucinations. How can we leverage the strengths of both to mitigate their respectiveweaknesses? We present a proof of concept that integrates Neo4j with LLM-basedagents, demonstrating their synergistic potential in the realm of biomedical datascience and research. By combining the structured, reliable nature of knowledgegraphs with the generative and summarizing capabilities of LLMs, we aim to create arobust framework for more effective data analysis and insight generation.