Evidence-Graded vs. Flat Knowledge Graphs: A Comparative Analysis of Trust Representation Approaches

Key Points

• This research aims to compare evidence-graded knowledge graphs with traditional flat knowledge graphs to assess their representational capabilities.
• Theoretical comparison of evidence-graded KGs against Wikidata, DBpedia, NELL, and Google Knowledge Vault.
• Evaluation based on seven dimensions: granularity, calibration, temporal awareness, and others.
• Inclusion of domain-specific case studies across medical, legal, and financial fields.
• Evidence-graded KGs show enhanced interpretability and granularity compared to flat graphs.
• Trade-offs identified in terms of cost-benefit for different applications.
• Temporal awareness and source attribution were found to vary significantly across platforms.

Abstract