Don’t Miss the Labels: Label-semantic Augmented Meta-Learner for Few-Shot Text Classification

Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, pages 2773–2782 August 1–6, 2021. ©2021 Association for Computational Linguistics 2773 Don’t Miss the Labels: Label-semantic Augmented Meta-Learner for Few-Shot Text Classification Qiaoyang Luo, Lingqiao Liu∗, Yuhao Lin and Wei Zhang The Univeristy of Adelaide {qiaoyang.luo, lingqiao.liu, wei.e.zhang}@adelaide.edu.au linyuhao0514@gmail.com Abstract Increasing studies leverage pre-trained lan- guage models and meta-learning frameworks to solve few-shot text classification problems. Most of the current studies focus on building a meta-learner from the information of input texts but ignore abundant semantic informa- tion beneath class labels. In this work, we show that class-label information can be uti- lized for extracting more discriminative fea- ture representation of the input text from a pre- trained language model like BERT, and can achieve a performance boost when the sam- ples are scarce. Building on top of this dis- covery, we propose a framework called Label- semantic augmented meta-learner (LaSAML) to make full use of label semantics. We sys- tematically investigate various factors in this framework and show that it can be plugged into the existing few-shot text classification system. Through extensive experiments, we demonstrate that the few-shot text classifica- tion system upgraded by LaSAML can lead to significant performance improvement over its original counterparts. 1 Introduction The remarkable capability of quickly learning new concepts from a few training samples is one of the advantages of the human learning system over the current machine learning system. Motivated by this gap, research in few-shot learning has received in- creasing attention in the past decade. Meta-learning (Vinyals et al., 2016; Snell et al., 2017; Finn et al., 2017), as the dominant methodology in few-shot learning, tackles the problem by learning a map- ping function from a few support samples to a classifier through a meta-training dataset. Most existing meta-learning systems (Snell et al., 2017; Sung et al., 2018) were developed or at least evalu- ated in the field of computer vision. More recently, ∗Corresponding Author Figure 1: The example of fine-grained intent queries from two domains in Clinc150 dataset. Class 1 and Class 2 refer to PAY BILL and BILL DUE. Class 3 and Class 4 represent TRAVEL SUGGESTION and TRAVEL ALERT. few-shot learning has been introduced to the NLP field and in particular, text classification (Yu et al., 2018; Geng et al., 2019), as it is the fundamental task in natural language understanding. In parallel to few-shot learning, pre-trained language models (PLMs) (Devlin et al., 2019; Radford et al., 2019) have revolutionized the NLP fields and show strong evidence of being able to perform well in low data regime when transferred to downstream tasks. Despite the impressive progress of meta-learning and PLMs, however, most existing few-shot clas- sification systems (Geng et al., 2019; Bao et al., 2020) ignore an important information source — semantic of class labels. When the number of train- ing samples is limited, merely using the input texts per class can lead to ambiguity in interpreting the definition of class. Considering the two groups of examples in Figure …

Content truncated to 500 words