Answers are grounded in this paper and cite their sources.
This approach evaluates knowledge credibility in domain-specific contexts, suggesting effective methods for increasing reliability.
Abstract Knowledge graphs have become pivotal infrastructure in training large language models (LLMs), powering knowledge question-answering (QA) systems, and enabling semantic search applications. However, the construction and deployment of domain-specific knowledge graphs face critical challenges: how to ensure the reliability of knowledge derived from these graphs, or more fundamentally, how to establish the credibility of encoded knowledge. While credibility concepts have been applied to specific technical tasks like knowledge graph completion and error detection, systematic methodologies for assessing knowledge graph credibility remain severely understudied. To fill this gap, we propose a multidimensional approach in this paper for assessing knowledge credibility in domain-specific knowledge graphs through three orthogonal dimensions: entity-relationship, attribute value, and problem-solving. The case study shows that the approach proposed in this paper can effectively distinguish between trusted and untrusted knowledge with high accuracy, accuracy, and recall rate.
No takes yet. Share an insight, caveat, or question.
Yin et al. (2025) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: