Contrastive Bootstrapping for Label Refinement
This addresses the need for accurate fine-grained categorization in real-world text services where new categories emerge periodically, but it is incremental as it builds on existing coarse-to-fine mapping approaches.
The paper tackles the problem of fine-grained text classification using only coarse-grained labels and a coarse-to-fine mapping, proposing a contrastive clustering-based bootstrapping method that iteratively refines labels, and it outperforms state-of-the-art methods by a large margin on NYT and 20News datasets.
Traditional text classification typically categorizes texts into pre-defined coarse-grained classes, from which the produced models cannot handle the real-world scenario where finer categories emerge periodically for accurate services. In this work, we investigate the setting where fine-grained classification is done only using the annotation of coarse-grained categories and the coarse-to-fine mapping. We propose a lightweight contrastive clustering-based bootstrapping method to iteratively refine the labels of passages. During clustering, it pulls away negative passage-prototype pairs under the guidance of the mapping from both global and local perspectives. Experiments on NYT and 20News show that our method outperforms the state-of-the-art methods by a large margin.