arXiv 2306.083022024-01-25cs.CLcs.AI

统一大语言模型与知识图谱：路线图

Unifying Large Language Models and Knowledge Graphs: A Roadmap

Griffith University（格里菲斯大学）；Monash University（莫纳什大学）；Nanyang Technological University（南洋理工大学）；Beijing University of Technology（北京工业大学）；Hefei University of Technology（合肥工业大学）；Zhejiang Lab（之江实验室）

Shirui Pan, Linhao Luo, Yufei Wang, Chen Chen, Jiapu Wang, Xindong Wu

备注：: A short version of this paper was accepted by IEEE Transactions on Knowledge and Data Engineering (TKDE)
发表信息：: IEEE Transactions on Knowledge and Data Engineering (TKDE) 2024
DOI：: 10.1109/TKDE.2024.3352100

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出统一大语言模型与知识图谱的三框架路线图，涵盖 KG 增强 LLM、LLM 增强 KG 及二者协同，并综述现有进展与未来方向。

摘要：

大语言模型（LLM），例如 ChatGPT 和 GPT4，凭借其涌现能力和泛化能力，正在自然语言处理与人工智能领域掀起新的浪潮。然而，LLM 是黑盒模型，往往难以捕获和访问事实性知识。相比之下，知识图谱（KG）——例如 Wikipedia 和 Huapu——是显式存储丰富事实性知识的结构化知识模型。KG 可以通过为推理和可解释性提供外部知识来增强 LLM。同时，KG 本身难以构建且处于不断演化之中，这使得现有 KG 方法在生成新事实和表示未见知识方面面临挑战。因此，将 LLM 与 KG 统一起来并同时利用二者优势具有互补意义。本文提出了一个面向未来的 LLM 与 KG 统一路线图。该路线图包含三个通用框架：1）KG 增强的 LLM，在 LLM 的预训练和推理阶段融入 KG，或用于增强对 LLM 已学知识的理解；2）LLM 增强的 KG，利用 LLM 完成嵌入、补全、构建、图到文本生成和问答等不同 KG 任务；3）协同的 LLM + KG，其中 LLM 与 KG 地位平等，并以互利方式协作，在数据与知识共同驱动的双向推理中同时增强 LLM 和 KG。我们回顾并总结了路线图中三个框架下的现有工作，并指明其未来研究方向。

英文摘要：

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 evolving 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; and 3) 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.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。