Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Label-Guided Prompt for Multi-label Few-shot Aspect Category Detection

ChaoFeng Guan, YaoHui Zhu, Yu Bai, LingYun Wang

TL;DR

This work designs label-specific prompts to represent sentences by combining crucial contextual and semantic information in a label-guided prompt method to represent sentences and categories in multi-label few-shot aspect category detection.

Abstract

Multi-label few-shot aspect category detection aims at identifying multiple aspect categories from sentences with a limited number of training instances. The representation of sentences and categories is a key issue in this task. Most of current methods extract keywords for the sentence representations and the category representations. Sentences often contain many category-independent words, which leads to suboptimal performance of keyword-based methods. Instead of directly extracting keywords, we propose a label-guided prompt method to represent sentences and categories. To be specific, we design label-specific prompts to represent sentences by combining crucial contextual and semantic information. Further, the label is introduced into a prompt to obtain category descriptions by utilizing a large language model. This kind of category descriptions contain the characteristics of the aspect categories, guiding the construction of discriminative category prototypes. Experimental results on two public datasets show that our method outperforms current state-of-the-art methods with a 3.86% - 4.75% improvement in the Macro-F1 score.

Label-Guided Prompt for Multi-label Few-shot Aspect Category Detection

TL;DR

This work designs label-specific prompts to represent sentences by combining crucial contextual and semantic information in a label-guided prompt method to represent sentences and categories in multi-label few-shot aspect category detection.

Abstract

Multi-label few-shot aspect category detection aims at identifying multiple aspect categories from sentences with a limited number of training instances. The representation of sentences and categories is a key issue in this task. Most of current methods extract keywords for the sentence representations and the category representations. Sentences often contain many category-independent words, which leads to suboptimal performance of keyword-based methods. Instead of directly extracting keywords, we propose a label-guided prompt method to represent sentences and categories. To be specific, we design label-specific prompts to represent sentences by combining crucial contextual and semantic information. Further, the label is introduced into a prompt to obtain category descriptions by utilizing a large language model. This kind of category descriptions contain the characteristics of the aspect categories, guiding the construction of discriminative category prototypes. Experimental results on two public datasets show that our method outperforms current state-of-the-art methods with a 3.86% - 4.75% improvement in the Macro-F1 score.
Paper Structure (15 sections, 9 equations, 3 figures, 7 tables)

This paper contains 15 sections, 9 equations, 3 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Multi-label Few-shot Aspect Category Detection
  4. Prompt Learning
  5. Methods
  6. Problem Formulation
  7. Prompt Enhanced Sentence Representation
  8. Prompt Enhanced Prototype Generation
  9. Training and Inference
  10. Experiments
  11. Datasets and Experimental Setups
  12. Experimental Results and Analysis
  13. Ablation Study
  14. Discussions
  15. Conclusions and Further Works

Figures (3)

  • Figure 1: The overview of our proposed LGP framework. The left part describes the sentences of support set and query set, with different colors representing different categories. The middle section describes the details of PESR and PEPG. PESR can enhance sentence representation by the prompt. PEPG guides the prototype generation through features of category descriptions.
  • Figure 2: Feature visualization of category prototype on FewAsp(Multi). The category prototypes are calculated from 3000 5-way 5-shot tasks. Different colors means different categories.
  • Figure 3: The impact on the number of tokens, token state and encoder parameters.