A method for extracting aquatic animal disease prevention and control events integrated with capsule network

Addressing the issue of long-tail event entity recognition in aquatic animal disease prevention and control, this paper proposes an event extraction method that integrates capsule networks. The method designs two parallel networks: the first utilizes BERT + TextCNN to extract initial and local features from the text, while Multi-BiLSTM further captures multi-dimensional dependency information features. The second network employs capsule networks to extract local features and learns spatial semantic relationships among different event entities. The features extracted from both networks are then fused. Experimental results demonstrate that this method achieves significant recognition performance on the aquatic animal disease prevention and control event dataset, with an F1 score of 75.83%, effectively addressing the challenge of long-tail event entity recognition. Capsule networks elevate aquatic disease event recognition to 75.83%, advancing aquaculture health monitoring. The BTCapMB framework offers a novel approach to extract complex event entities, vital for proactive disease management in aquaculture. Precise extraction of long-tail entities enables timely interventions, mitigating economic losses and antibiotic demand, fostering sustainable aquaculture practices.