Hierarchy-Aware and Label Balanced Model for Hierarchical Text Classification

Jun Zhang, Yubin Li, Fanfan Shen, Chenxi Xia, Hai Tan, Yanxiang He

Abstract: Highlights•Propose a hierarchy-aware and label balanced model based on the model HGCLR for hierarchical text classification.•Propose a new method called multi-label negative supervision to obtain more hierarchy-aware text representations.•Adopt asymmetric loss instead of cross-entropy loss as classification loss to deal with label imbalance.