Domain knowledge-powered attention for air traffic management hazardous events classification

Ziyi Guo, Weili Zeng, Zhibin Quan, Xianghua Tan

Abstract: Highlights•A deep learning model for hazardous events text classification is proposed.•Designed attention mechanism incorporates simply structed textual domain knowledge.•Trainable modules facilitate two-tier feature extraction.•Model is evaluated on a multi-label Chinese database with 21 classes.