Document-Level Chemical-Induced Disease Semantic Relation Extraction Using Bidirectional Long Short-Term Memory on Dependency Graph

Abstract Identifying chemical-induced disease (CID) semantic relations in the biomedical literature, including both intra- and inter-sentence interactions, has significant implications for various downstream applications. Although advanced methods have been proposed, they often overlook cross-sentence dependency information, which is crucial accurately predicting relations. In this study, we propose DEGREx, a novel graph-based neural model that presents document as graph. DEGREx improves long-distance relation extraction by allowing direct information exchange among graph nodes through connections. The transition process based on idea of controller gates long short-term memory networks. Our model, exerts multi-task learning framework to jointly train with named entity recognition, improving performance CID task. Experimental results benchmark dataset demonstrate our outperforms all nine compared recent state-of-the-art models.