Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Heuristic-enhanced Candidates Selection strategy for GPTs tackle Few-Shot Aspect-Based Sentiment Analysis

Baoxing Jiang, Yujie Wan, Shenggen Ju

TL;DR

This work tackles FSABSA under data-scarce conditions by introducing a two-stage AiO framework that couples a PLM-based backbone for generating heuristic candidates with GPT-based inference guided by task-specific prompts. The Heuristic-Enhanced Candidates Selection (HCS) strategy, together with a multi-shot ensemble (T × K), enables GPTs to deliver precise predictions across the AOPE, ALSC, and ASTE sub-tasks, outperforming direct GPT usage and several baselines on five datasets. The study provides a reproducible learning paradigm for applying GPTs to FSABSA, demonstrates robust generalization across domains, and highlights the practical value of combining structured prompts with context-aware candidate selection. While effective, the approach incurs computational costs for candidate generation and prompt construction, indicating avenues for efficiency improvements and broader GPT evaluation in future work.

Abstract

Few-Shot Aspect-Based Sentiment Analysis (FSABSA) is an indispensable and highly challenging task in natural language processing. However, methods based on Pre-trained Language Models (PLMs) struggle to accommodate multiple sub-tasks, and methods based on Generative Pre-trained Transformers (GPTs) perform poorly. To address the above issues, the paper designs a Heuristic-enhanced Candidates Selection (HCS) strategy and further proposes All in One (AiO) model based on it. The model works in a two-stage, which simultaneously accommodates the accuracy of PLMs and the generalization capability of GPTs. Specifically, in the first stage, a backbone model based on PLMs generates rough heuristic candidates for the input sentence. In the second stage, AiO leverages LLMs' contextual learning capabilities to generate precise predictions. The study conducted comprehensive comparative and ablation experiments on five benchmark datasets. The experimental results demonstrate that the proposed model can better adapt to multiple sub-tasks, and also outperforms the methods that directly utilize GPTs.

Heuristic-enhanced Candidates Selection strategy for GPTs tackle Few-Shot Aspect-Based Sentiment Analysis

TL;DR

This work tackles FSABSA under data-scarce conditions by introducing a two-stage AiO framework that couples a PLM-based backbone for generating heuristic candidates with GPT-based inference guided by task-specific prompts. The Heuristic-Enhanced Candidates Selection (HCS) strategy, together with a multi-shot ensemble (T × K), enables GPTs to deliver precise predictions across the AOPE, ALSC, and ASTE sub-tasks, outperforming direct GPT usage and several baselines on five datasets. The study provides a reproducible learning paradigm for applying GPTs to FSABSA, demonstrates robust generalization across domains, and highlights the practical value of combining structured prompts with context-aware candidate selection. While effective, the approach incurs computational costs for candidate generation and prompt construction, indicating avenues for efficiency improvements and broader GPT evaluation in future work.

Abstract

Few-Shot Aspect-Based Sentiment Analysis (FSABSA) is an indispensable and highly challenging task in natural language processing. However, methods based on Pre-trained Language Models (PLMs) struggle to accommodate multiple sub-tasks, and methods based on Generative Pre-trained Transformers (GPTs) perform poorly. To address the above issues, the paper designs a Heuristic-enhanced Candidates Selection (HCS) strategy and further proposes All in One (AiO) model based on it. The model works in a two-stage, which simultaneously accommodates the accuracy of PLMs and the generalization capability of GPTs. Specifically, in the first stage, a backbone model based on PLMs generates rough heuristic candidates for the input sentence. In the second stage, AiO leverages LLMs' contextual learning capabilities to generate precise predictions. The study conducted comprehensive comparative and ablation experiments on five benchmark datasets. The experimental results demonstrate that the proposed model can better adapt to multiple sub-tasks, and also outperforms the methods that directly utilize GPTs.
Paper Structure (18 sections, 9 equations, 4 figures, 9 tables)

This paper contains 18 sections, 9 equations, 4 figures, 9 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Approach
  4. GPT In-context Learning
  5. GPTs for FSABSA
  6. The AiO Framework
  7. Stage 1. Heuristic-Enhanced Candidates Selection
  8. Stage 2. Answer Inference
  9. Experiment Details
  10. Hyperparameter Settings
  11. Datasets
  12. Experiment Results
  13. Compare with Baselines
  14. Ablation Studies
  15. Impact of Shot Size
  16. ...and 3 more sections

Figures (4)

  • Figure 1: Example sentence with three entities where aspects, opinions, and sentiments are marked in different colors.
  • Figure 2: Overview of AiO model. The left part is heuristic-enhanced candidates selection stage and the right part is answer inference stage.
  • Figure 3: Example of constructing a prompt template based on a particular sub-task $\varsigma$.
  • Figure 4: Effect of shot size on results. Plotted as an example of ChatGPT-3.5 and GPT-J completing the ASTE task on top of RES16, with fitted curves set up for each fold and fitted extremes labeled.