Distinct Label Representations for Few-Shot Text Classification

Abstract

Few-shot text classification aims to classify inputs whose label has only a few examples.Previous studies overlooked the semantic relevance between label representations. Therefore, they are easily confused by labels that are semantically relevant. To address this problem, we propose a method that generates distinct label representations that embed information specific to each label. Our method is widely applicable to conventional few-shot classification models. Experimental results show that our method significantly improved the performance of few-shot text classification across models and datasets.

Improving Few-Shot Relation Classification by Prototypical Representation Learning with Definition Text

“…Two types of efforts have been devoted to improving few-shot RC. Firstly, some approaches (Ye and Ling, 2019;Gao et al, 2019a;Han et al, 2021;Ren et al, 2020;Ohashi et al, 2021) design specific model architectures such as using attention mechanism to model complex interactions between labeled instances. However, these approaches are still limited when the few labeled instances are atypical and does not reflect the general patterns of the relation.…”