Label-Aware Hierarchical Ranking Model for Multi-Label Text Classification

Multi-label text classification involves assigning multiple relevant categories to a single text, enabling applications in academic indexing, medical diagnostics, and e-commerce. However, existing models often fail to capture complex text-label relationships and lack robust mechanisms for ranking label relevance, limiting their effectiveness. This paper proposes the Label-Aware Hierarchical Ranking Model, a novel approach that combines contextual embeddings, custom attention mechanisms, and gradient boosting to enhance label ranking. The proposed model integrates text and label embeddings with a contextual similarity attention module to capture text-label relationships and inter-label dependencies, followed by a hierarchical ranking mechanism for refining label prioritization. Experiments conducted on both the ArXiv Academic Paper dataset and the Reuters-21578 dataset demonstrate that the proposed model surpasses state-of-the-art models, achieving Precision@1 scores of 86.90 % and 94.47 %, respectively. These results advance multi-label text classification methodologies, offering more accurate, context-aware, and practically applicable classifications.