LLMs for Knowledge Graph Construction and Reasoning: Recent Capabilities and Future Opportunities

2023-05-22

World wide web (Bussum)

AI Summary

This work systematically evaluates large language models (LLMs) for knowledge graph (KG) construction and reasoning across four core tasks: entity/relation extraction, event identification, link prediction, and question answering. The study identifies a key insight: LLMs serve more effectively as *reasoning assistants* than few-shot extractors. To advance evaluation, the authors introduce the novel *virtual knowledge extraction* task and the VINE benchmark dataset. They further propose AutoKG—a multi-agent framework integrating prompt engineering, external knowledge retrieval, and dynamic verification—to enable end-to-end KG construction and reasoning. Experiments demonstrate that GPT-4–based AutoKG outperforms fine-tuned models on multiple reasoning benchmarks and achieves superior cross-domain generalization. The codebase and VINE dataset are publicly released to foster deeper integration of KGs and LLMs.

Abstract

This paper presents an exhaustive quantitative and qualitative evaluation of Large Language Models (LLMs) for Knowledge Graph (KG) construction and reasoning. We engage in experiments across eight diverse datasets, focusing on four representative tasks encompassing entity and relation extraction, event extraction, link prediction, and question-answering, thereby thoroughly exploring LLMs' performance in the domain of construction and inference. Empirically, our findings suggest that LLMs, represented by GPT-4, are more suited as inference assistants rather than few-shot information extractors. Specifically, while GPT-4 exhibits good performance in tasks related to KG construction, it excels further in reasoning tasks, surpassing fine-tuned models in certain cases. Moreover, our investigation extends to the potential generalization ability of LLMs for information extraction, leading to the proposition of a Virtual Knowledge Extraction task and the development of the corresponding VINE dataset. Based on these empirical findings, we further propose AutoKG, a multi-agent-based approach employing LLMs and external sources for KG construction and reasoning. We anticipate that this research can provide invaluable insights for future undertakings in the field of knowledge graphs. The code and datasets are in https://github.com/zjunlp/AutoKG.

Problem

• Large Language Models
• Knowledge Graph Tasks
• Few-shot Learning

Innovation

• Large Language Models
• Knowledge Graph Construction
• AutoKG Method