Unifying Large Language Models and Knowledge Graphs: A Roadmap

Объединение больших языковых моделей и графов знаний: дорожная карта
Xindong Wu, Shirui Pan, Linhao Luo, Yufei Wang, Chen Chen, Jiapu Wang
2024-01-10

KG-enhanced LLMsLLM-augmented KGsbidirectional reasoningknowledge graphslarge language models
Large language models (LLMs), such as ChatGPT and GPT4, are making new waves in the field of natural language processing and artificial intelligence, due to their emergent ability and generalizability. However, LLMs are black-box models, which often fall short of capturing and accessing factual knowledge. In contrast, Knowledge Graphs (KGs), Wikipedia and Huapu for example, are structured knowledge models that explicitly store rich factual knowledge. KGs can enhance LLMs by providing external knowledge for inference and interpretability. Meanwhile, KGs are difficult to construct and evolve by nature, which challenges the existing methods in KGs to generate new facts and represent unseen knowledge. Therefore, it is complementary to unify LLMs and KGs together and simultaneously leverage their advantages. In this article, we present a forward-looking roadmap for the unification of LLMs and KGs. Our roadmap consists of three general frameworks, namely,1) KG-enhanced LLMs,which incorporate KGs during the pre-training and inference phases of LLMs, or for the purpose of enhancing understanding of the knowledge learned by LLMs;2) LLM-augmented KGs,that leverage LLMs for different KG tasks such as embedding, completion, construction, graph-to-text generation, and question answering; and3) Synergized LLMs + KGs, in which LLMs and KGs play equal roles and work in a mutually beneficial way to enhance both LLMs and KGs for bidirectional reasoning driven by both data and knowledge. We review and summarize existing efforts within these three frameworks in our roadmap and pinpoint their future research directions.
1
A mutually beneficial LLM–KG integration is proposed as a basis for bidirectional reasoning driven jointly by data and structured knowledge, with future research directions identified.
2
Knowledge graphs offer structured, explicit factual knowledge that can improve LLM inference and interpretability, but are difficult to construct and continuously evolve.
3
LLMs can support KG embedding, completion, construction, graph-to-text generation, and question answering, addressing challenges in generating and representing new knowledge.
4
Large language models provide strong generalization but remain black-box systems that often struggle to capture and access factual knowledge.
5
The paper organizes LLM–KG integration into three frameworks: KG-enhanced LLMs, LLM-augmented KGs, and synergized LLMs plus KGs.

the unification of large language models (LLMs) and knowledge graphs (KGs)

frameworks, methods, and future research directions for mutually leveraging LLMs and KGs to improve factual knowledge access, interpretability, knowledge construction, completion, and bidirectional reasoning

Publication Details
Publication Date
2024-01-10
Journal
Publisher
ISSN
Cited by
2567
Access Type
Author Information
Authors
Xindong Wu
Shirui Pan
Linhao Luo
Yufei Wang
Chen Chen
Jiapu Wang
Explore further
Open the scid.ai AI chat with a ready-made request: it will find papers on a similar topic and help build a literature review.
Find similar papers in the chat →
Make a presentation
100%