Label-Aware Document Representation via Hybrid Attention for Extreme Multi-Label Text Classification

Extreme multi-label text classification (XMTC) aims at tagging a document with most relevant labels from an extremely large-scale label set. It is challenging problem especially for the tail because there are only few training documents to build classifier. This paper motivated better explore semantic relationship between each and extreme by taking advantage of both content correlation. Our objective establish explicit label-aware representation hybrid attention deep neural network model(LAHA). LAHA consists three parts. The first part adopts self-attention mechanism detect contribution word labels. second exploits structure determine connection words in same latent space. An adaptive fusion strategy designed third obtain final so that essence previous two parts can be sufficiently integrated. Extensive experiments have been conducted on six benchmark datasets comparing state-of-the-art methods. results show superiority our proposed method,