Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Enhanced Coherence-Aware Network with Hierarchical Disentanglement for Aspect-Category Sentiment Analysis

Jin Cui, Fumiyo Fukumoto, Xinfeng Wang, Yoshimi Suzuki, Jiyi Li, Noriko Tomuro, Wanzeng Kong

TL;DR

The paper addresses aspect-category sentiment analysis (ACSA) by modeling document-wide coherence and disentangling category- and sentiment-specific information within reviews. It introduces ECAN, an enhanced coherence-aware network that combines XLNet-based coherence modeling, hierarchical disentanglement for category and sentiment blocks, word-level syntactic learning via dependency graphs, and a multi-task learning framework with contrastive and supervision losses. Empirical results across four benchmark datasets show state-of-the-art performance in both aspect-category detection (ACD) and aspect-category sentiment classification (ACSC), with strong ablations highlighting the contributions of coherence modeling, disentanglement, and syntax. The work advances ACSA by enabling more precise, interpretable separation of category and sentiment cues in context, with practical implications for more accurate and scalable opinion mining.

Abstract

Aspect-category-based sentiment analysis (ACSA), which aims to identify aspect categories and predict their sentiments has been intensively studied due to its wide range of NLP applications. Most approaches mainly utilize intrasentential features. However, a review often includes multiple different aspect categories, and some of them do not explicitly appear in the review. Even in a sentence, there is more than one aspect category with its sentiments, and they are entangled intra-sentence, which makes the model fail to discriminately preserve all sentiment characteristics. In this paper, we propose an enhanced coherence-aware network with hierarchical disentanglement (ECAN) for ACSA tasks. Specifically, we explore coherence modeling to capture the contexts across the whole review and to help the implicit aspect and sentiment identification. To address the issue of multiple aspect categories and sentiment entanglement, we propose a hierarchical disentanglement module to extract distinct categories and sentiment features. Extensive experimental and visualization results show that our ECAN effectively decouples multiple categories and sentiments entangled in the coherence representations and achieves state-of-the-art (SOTA) performance. Our codes and data are available online: \url{https://github.com/cuijin-23/ECAN}.

Enhanced Coherence-Aware Network with Hierarchical Disentanglement for Aspect-Category Sentiment Analysis

TL;DR

The paper addresses aspect-category sentiment analysis (ACSA) by modeling document-wide coherence and disentangling category- and sentiment-specific information within reviews. It introduces ECAN, an enhanced coherence-aware network that combines XLNet-based coherence modeling, hierarchical disentanglement for category and sentiment blocks, word-level syntactic learning via dependency graphs, and a multi-task learning framework with contrastive and supervision losses. Empirical results across four benchmark datasets show state-of-the-art performance in both aspect-category detection (ACD) and aspect-category sentiment classification (ACSC), with strong ablations highlighting the contributions of coherence modeling, disentanglement, and syntax. The work advances ACSA by enabling more precise, interpretable separation of category and sentiment cues in context, with practical implications for more accurate and scalable opinion mining.

Abstract

Aspect-category-based sentiment analysis (ACSA), which aims to identify aspect categories and predict their sentiments has been intensively studied due to its wide range of NLP applications. Most approaches mainly utilize intrasentential features. However, a review often includes multiple different aspect categories, and some of them do not explicitly appear in the review. Even in a sentence, there is more than one aspect category with its sentiments, and they are entangled intra-sentence, which makes the model fail to discriminately preserve all sentiment characteristics. In this paper, we propose an enhanced coherence-aware network with hierarchical disentanglement (ECAN) for ACSA tasks. Specifically, we explore coherence modeling to capture the contexts across the whole review and to help the implicit aspect and sentiment identification. To address the issue of multiple aspect categories and sentiment entanglement, we propose a hierarchical disentanglement module to extract distinct categories and sentiment features. Extensive experimental and visualization results show that our ECAN effectively decouples multiple categories and sentiments entangled in the coherence representations and achieves state-of-the-art (SOTA) performance. Our codes and data are available online: \url{https://github.com/cuijin-23/ECAN}.
Paper Structure (26 sections, 15 equations, 4 figures, 5 tables)

This paper contains 26 sections, 15 equations, 4 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Task Definition
  4. Approach
  5. Coherence-Aware Representation Learning with XLNet
  6. Hierarchical Disentanglement
  7. Category Disentanglement
  8. Sentiment Disentanglement
  9. Word-Level Syntactic Learning for Enhancing Sentiment Semantics
  10. Multi-Task Learning
  11. Experiments
  12. Datasets and Evaluation Metrics
  13. Baselines
  14. Implementation Details
  15. Main Results
  16. ...and 11 more sections

Figures (4)

  • Figure 1: A review (ID: P#9) from SemEval-2015: Words with angle brackets and the colored circle show aspect categories. "$+$," "$-$," and "$\circ$" denote positive, negative, and neutral sentiments, respectively. Arcs indicate that they have an identical category or sentiment in a coherent review.
  • Figure 2: The main framework of our proposed method. It consists of four modules: (1) coherence-aware representation learning, (2) hierarchical disentanglement, (3) word-level syntactic learning, and (4) multi-task learning.
  • Figure 3: The visualization result of category and sentiment disentanglement. The top box on the left/right side indicates a sentence in a review with the source and disentangled categories and sentiments by the ECAN. The blue color and the red color denote the visualized correlations between categories and between sentiments in disentanglements. The warmer the tone colors, the higher the correlations.
  • Figure 4: The effect of hyperparameters $d_c$ and $d_s$.