Large Language Models and Data Quality for Knowledge Graphs

Abstract

Knowledge Graphs (KGs) have become essential for applications such as virtual assistants, web search, reasoning, and information access and management. Prominent examples include Wikidata, DBpedia, YAGO, and NELL, which large companies widely use for structuring and integrating data. Constructing KGs involves various AI-driven processes, including data integration, entity recognition, relation extraction, and active learning. However, automated methods often lead to sparsity and inaccuracies, making rigorous KG quality evaluation crucial for improving construction methodologies and ensuring reliable downstream applications. Despite its importance, large-scale KG quality assessment remains an underexplored research area.

The rise of Large Language Models (LLMs) introduces both opportunities and challenges for KG construction and evaluation. LLMs can enhance contextual understanding and reasoning in KG systems but also pose risks, such as introducing misinformation or “hallucinations” that could degrade KG integrity. Effectively integrating LLMs into KG workflows requires robust quality control mechanisms to manage errors and ensure trustworthiness.

This special issue explores the intersection of KGs and LLMs, emphasizing human–machine collaboration for KG construction and evaluation. We present contributions on LLM-assisted KG generation, large-scale KG quality assessment, and quality control mechanisms for mitigating LLM-induced errors. Topics covered include KG construction methodologies, LLM deployment in KG systems, scalable KG evaluation, human-in-the-loop approaches, domain-specific applications, and industrial KG maintenance. By advancing research in these areas, this issue fosters innovation at the convergence of KGs and LLMs.