Scholarlink

Dialogue Aspect-based Sentiment Quadruple Analysis (DiaASQ) is a task aimed at extracting target-aspect-opinion-sentiment quadruples from dialogues, with research covering multiple dimensions such as span matching, pair extraction, and quadruple recognition. Existing methods for the DiaASQ task face three main limitations: (1) Chaotic dialogue matching, (2) Single long-range dependency capture, and (3) Insufficient understanding of the dynamic changes in individual sentiments. To overcome these limitations, we propose a novel Dialogue Multi-dimensional Feature Dividing and Fusion (DiaMFDF) model. This model effectively divides the features of different dimensions in the dialogue and intelligently integrates them to enhance the performance of DiaASQ. We conduct experiments on two baseline datasets, and the results show that DiaMFDF achieves SOTA performance.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Dialogue Multi-dimensional Feature Dividing and Fusion Model for Dialogue Aspect-Based Sentiment Quadruple Analysis

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ying Ding,
  • Yanxu Mao,
  • Shunli Zhang

摘要

Dialogue Aspect-based Sentiment Quadruple Analysis (DiaASQ) is a task aimed at extracting target-aspect-opinion-sentiment quadruples from dialogues, with research covering multiple dimensions such as span matching, pair extraction, and quadruple recognition. Existing methods for the DiaASQ task face three main limitations: (1) Chaotic dialogue matching, (2) Single long-range dependency capture, and (3) Insufficient understanding of the dynamic changes in individual sentiments. To overcome these limitations, we propose a novel Dialogue Multi-dimensional Feature Dividing and Fusion (DiaMFDF) model. This model effectively divides the features of different dimensions in the dialogue and intelligently integrates them to enhance the performance of DiaASQ. We conduct experiments on two baseline datasets, and the results show that DiaMFDF achieves SOTA performance.