Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Let's Rectify Step by Step: Improving Aspect-based Sentiment Analysis with Diffusion Models

Shunyu Liu, Jie Zhou, Qunxi Zhu, Qin Chen, Qingchun Bai, Jun Xiao, Liang He

TL;DR

This work tackles the challenging problem of precise aspect boundary extraction in ABSA by introducing DiffusionABSA, a diffusion-model framework that progressively refines aspect terms through a forward corruption and backward denoising process. A syntax-aware temporal attention (SynTA) module guides boundary estimation by integrating POS and dependency information with temporal cues, while an end-to-end network predicts start/end spans and sentiment labels. Empirical results across eight ABSA datasets show competitive or state-of-the-art performance, with notable gains for longer aspects and robust ablations confirming the value of SynTA and diffusion-based refinement. The approach demonstrates the potential of controlled generation in token-level ABSA and offers a solid platform for extending to more complex ABSA tasks in the future.

Abstract

Aspect-Based Sentiment Analysis (ABSA) stands as a crucial task in predicting the sentiment polarity associated with identified aspects within text. However, a notable challenge in ABSA lies in precisely determining the aspects' boundaries (start and end indices), especially for long ones, due to users' colloquial expressions. We propose DiffusionABSA, a novel diffusion model tailored for ABSA, which extracts the aspects progressively step by step. Particularly, DiffusionABSA gradually adds noise to the aspect terms in the training process, subsequently learning a denoising process that progressively restores these terms in a reverse manner. To estimate the boundaries, we design a denoising neural network enhanced by a syntax-aware temporal attention mechanism to chronologically capture the interplay between aspects and surrounding text. Empirical evaluations conducted on eight benchmark datasets underscore the compelling advantages offered by DiffusionABSA when compared against robust baseline models. Our code is publicly available at https://github.com/Qlb6x/DiffusionABSA.

Let's Rectify Step by Step: Improving Aspect-based Sentiment Analysis with Diffusion Models

TL;DR

This work tackles the challenging problem of precise aspect boundary extraction in ABSA by introducing DiffusionABSA, a diffusion-model framework that progressively refines aspect terms through a forward corruption and backward denoising process. A syntax-aware temporal attention (SynTA) module guides boundary estimation by integrating POS and dependency information with temporal cues, while an end-to-end network predicts start/end spans and sentiment labels. Empirical results across eight ABSA datasets show competitive or state-of-the-art performance, with notable gains for longer aspects and robust ablations confirming the value of SynTA and diffusion-based refinement. The approach demonstrates the potential of controlled generation in token-level ABSA and offers a solid platform for extending to more complex ABSA tasks in the future.

Abstract

Aspect-Based Sentiment Analysis (ABSA) stands as a crucial task in predicting the sentiment polarity associated with identified aspects within text. However, a notable challenge in ABSA lies in precisely determining the aspects' boundaries (start and end indices), especially for long ones, due to users' colloquial expressions. We propose DiffusionABSA, a novel diffusion model tailored for ABSA, which extracts the aspects progressively step by step. Particularly, DiffusionABSA gradually adds noise to the aspect terms in the training process, subsequently learning a denoising process that progressively restores these terms in a reverse manner. To estimate the boundaries, we design a denoising neural network enhanced by a syntax-aware temporal attention mechanism to chronologically capture the interplay between aspects and surrounding text. Empirical evaluations conducted on eight benchmark datasets underscore the compelling advantages offered by DiffusionABSA when compared against robust baseline models. Our code is publicly available at https://github.com/Qlb6x/DiffusionABSA.
Paper Structure (27 sections, 20 equations, 2 figures, 6 tables)

This paper contains 27 sections, 20 equations, 2 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Aspect-based Sentiment Analysis
  4. Diffusion Models
  5. Our Proposed Framework
  6. Aspect Diffusion
  7. Forward Corruption Process
  8. Backward Denoising Process
  9. Denoising Neural Network
  10. Contextual Encoder
  11. Syntax-aware Temporal Attention (SynTA)
  12. Aspect Extractor
  13. Sentiment Classifier
  14. Loss Function
  15. Experiment Setting
  16. ...and 12 more sections

Figures (2)

  • Figure 1: The framework of DiffusionABSA.
  • Figure 2: Further analysis of DiffusionABSA.