A Tree-structured Neural Network Model for Joint Extraction of Adverse Drug Events

Extracting adverse drug events (ADEs) is one of the most challenging tasks in biomedical field. However, the existing studies fail to fully utilize the syntactic information and capture the long-range dependencies between entities, limiting the performance of the task. To address this issue, this paper introduces a tree-structured neural network model for jointly extracting ADEs entities and their relationships. Results on the publicly available dataset show that our proposed model can achieve 93.34% and 87.40% F1 scores on entity recognition and relation extraction, respectively, outperforming the current best models by a large margin.