Confidence Prediction Based on Uncertain Knowledge Graph Structure Embedding

Abstract

Abstract The development of large-scale knowledge graphs (KGs) has given rise to uncertain relational facts, leading to research on uncertain knowledge graph (KG) embeddings. While various studies have been conducted on the task of uncertain KG embeddings, they often employ simplistic scoring functions based on the internal interaction information among triplets to fit confidence scores, neglecting the rich neighborhood information. In light of this, we propose a novel model UKGSE for uncertain KG embeddings that captures the subgraph structural features formed by the neighbors of triplets, aiming to predict confidence scores for triplets. To validate the effectiveness of our model, we conduct confidence prediction tasks on benchmark datasets. The experimental results indicate that the performance of our proposed model surpasses mainstream embedding methods.