LSA: Modeling Aspect Sentiment Coherency via Local Sentiment Aggregation

TL;DR

This work tackles the underexplored problem of aspect sentiment coherency in ABSC by introducing Local Sentiment Aggregation (LSA), a lightweight paradigm that builds a differential-weighted sentiment aggregation window around target aspects. By leveraging three aspect-feature representations (BERT-SPC, Local Context Focus, and Syntactical LC), and a Differential Weighted Aggregation mechanism, LSA captures local sentiment coherence and improves both cluster-level coherency and standard ABSC accuracy. Across five public ABSC datasets, LSA achieves state-of-the-art results, with notable gains in adversarial defense scenarios, and its modular design allows easy integration with existing models. The work also provides thorough ablations and case studies, supporting the effectiveness and robustness of the local coherence approach, and releases open-source code for future research and application.

Abstract

Aspect sentiment coherency is an intriguing yet underexplored topic in the field of aspect-based sentiment classification. This concept reflects the common pattern where adjacent aspects often share similar sentiments. Despite its prevalence, current studies have not fully recognized the potential of modeling aspect sentiment coherency, including its implications in adversarial defense. To model aspect sentiment coherency, we propose a novel local sentiment aggregation (LSA) paradigm based on constructing a differential-weighted sentiment aggregation window. We have rigorously evaluated our model through experiments, and the results affirm the proficiency of LSA in terms of aspect coherency prediction and aspect sentiment classification. For instance, it outperforms existing models and achieves state-of-the-art sentiment classification performance across five public datasets. Furthermore, we demonstrate the promising ability of LSA in ABSC adversarial defense, thanks to its sentiment coherency modeling. To encourage further exploration and application of this concept, we have made our code publicly accessible. This will provide researchers with a valuable tool to delve into sentiment coherency modeling in future research.

Figures (6)

• Figure 1: An example of aspect sentiment clusters and aspect sentiment coherency.
• Figure 2: The local sentiment aggregation paradigm based on LCF/LCFS, denoted as LSA$_{T}$ and LSA$_{S}$.
• Figure 3: Window padding strategies for different situations.
• Figure 4: Trajectory visualization of learnable weights in gradient-based sentiment aggregation window optimization.
• Figure 5: The local sentiment aggregation paradigm based on BERT-SPC, denoted as LSA$_{P}$. "SA" indicates the self-attention encoder.