Abstract
Knowledge Graphs (KGs) play a vital role in data-driven applications such as virtual assistants, recommendation systems, and semantic search. The accuracy and reliability of these applications depend on the quality of the underlying KGs, making their evaluation crucial. However, manually assessing the quality of large-scale KGs is prohibitively expensive, requiring efficient strategies that minimize human annotation while ensuring guarantees on the assessments. Despite its importance, this challenge has received limited attention. This seminar presents recent advances in KG quality evaluation, introducing advanced sampling techniques and robust statistical methods that enhance reliability and efficiency. Particular attention is paid to the choice of the statistical intervals used to quantify the uncertainties associated with sampling. The seminar also examines real-world applications and outlines current challenges and limitations, particularly in dynamic environments where KG contents evolve over time.