# ChatGPT is not Enough: Enhancing Large Language Models with Knowledge Graphs for Fact-aware Language Modeling

**Authors:** Linyao Yang, Hongyang Chen, Zhao Li, Xiao Ding, Xindong Wu  
**arXiv:** [2306.11489](https://arxiv.org/abs/2306.11489)  
**Date:** 2023-06-20  
**Summary page:** https://www.summarizepaper.com/en/paper/2306.11489v1/

## Key points

- ChatGPT, a large language model (LLM), has gained attention for its emergent abilities
- LLMs have the potential to replace structured knowledge bases like knowledge graphs (KGs)
- Researchers propose enhancing LLMs with KGs to incorporate explicit factual knowledge
- Existing studies on knowledge graph enhanced PLMs (KGPLMs) have shown effectiveness
- Authors propose developing knowledge graph-enhanced large language models (KGLLMs)
- KGLLMs aim to enhance factual reasoning ability and open up new research avenues
- The paper reviews existing studies on enhancing PLMs with KGs and introduces KGPLM applications
- The development of KGLLM is proposed as a solution to improve LLM's ability to recall facts and generate informed responses
- This advancement holds promise for advancing research in large language models and their practical applications.

## Summary

Recently, ChatGPT, a representative large language model (LLM), has gained considerable attention for its powerful emergent abilities. Some researchers believe that LLMs have the potential to replace structured knowledge bases like knowledge graphs (KGs) and function as parameterized knowledge bases. To address the limitation of LLMs in recalling facts while generating knowledge-grounded content, researchers have proposed enhancing data-driven pre-trained language models (PLMs) with KGs to incorporate explicit factual knowledge into PLMs. Existing studies on knowledge graph enhanced PLMs (KGPLMs) have explored this approach and demonstrated its effectiveness. Inspired by these studies, the authors of this paper propose enhancing LLMs with KGs by developing knowledge graph-enhanced large language models (KGLLMs). By incorporating KGs into LLMs, KGLLMs aim to enhance their factual reasoning ability and open up new avenues for research in this field. In summary, this paper reviews existing studies on enhancing PLMs with KGs and introduces existing KGPLM applications. It also proposes the development of KGLLM as a solution to improve LLM's ability to recall facts and generate more informed responses to user queries. This advancement holds promise for advancing research in large language models and their practical applications.