English

Fast Multi-Resolution Transformer Fine-tuning for Extreme Multi-label Text Classification

Machine Learning 2021-11-01 v2 Artificial Intelligence Information Retrieval Machine Learning

Abstract

Extreme multi-label text classification (XMC) seeks to find relevant labels from an extreme large label collection for a given text input. Many real-world applications can be formulated as XMC problems, such as recommendation systems, document tagging and semantic search. Recently, transformer based XMC methods, such as X-Transformer and LightXML, have shown significant improvement over other XMC methods. Despite leveraging pre-trained transformer models for text representation, the fine-tuning procedure of transformer models on large label space still has lengthy computational time even with powerful GPUs. In this paper, we propose a novel recursive approach, XR-Transformer to accelerate the procedure through recursively fine-tuning transformer models on a series of multi-resolution objectives related to the original XMC objective function. Empirical results show that XR-Transformer takes significantly less training time compared to other transformer-based XMC models while yielding better state-of-the-art results. In particular, on the public Amazon-3M dataset with 3 million labels, XR-Transformer is not only 20x faster than X-Transformer but also improves the Precision@1 from 51% to 54%.

Keywords

Cite

@article{arxiv.2110.00685,
  title  = {Fast Multi-Resolution Transformer Fine-tuning for Extreme Multi-label Text Classification},
  author = {Jiong Zhang and Wei-cheng Chang and Hsiang-fu Yu and Inderjit S. Dhillon},
  journal= {arXiv preprint arXiv:2110.00685},
  year   = {2021}
}
R2 v1 2026-06-24T06:34:09.121Z