A Novel Neural Network Architecture for Biomedical Knowledge Graph Verification
Part Of
PACIS 2024 Proceedings
ISBN (of the container)
9781958200124
Date Issued
2024
Author(s)
Abstract
Biomedical knowledge graphs (KGs) play a crucial role in biomedical research and clinical settings. However, large biomedical knowledge graphs are often generated through automated extraction techniques from biomedical literature and typically contain erroneous statements about biomedical entities and relationships. Utilizing such noisy knowledge graphs in downstream applications will hamper the validity of biomedical research studies or even lead to erroneous conclusions. This study aims to design an effective method for determining the correctness of triplets (in the form of head entity-relation-tail entity) in a biomedical knowledge graph. We propose a knowledge graph verification (KGV) method which consists of a knowledge graph embedding (KGE) training stage and a KG triplet classification model training stage. We also design three different modes for KG triplet classification: Independent mode, Shared mode, and Multitask Learning (MTL) mode. Using SemMedDB as the source knowledge graph to train a KGE model and a dataset that consists of 3,760 biomedical triplets annotated by a domain expert, we empirically evaluate the effectiveness of our proposed KGV method. Our experimental results suggest that the MTL mode generally outperforms the other two modes (Independent and Shared modes).
Event(s)
Pacific-Asia Conference on Information Systems, Hi Chi Minh City, Vietnam
Subjects
Biomedical knowledge graph
Deep learning
Knowledge graph embedding
Knowledge graph verification
Multi-task learning
Publisher
Association for Information Systems
Type
conference paper