Constraint-Aware Counterfactual Editing for Aspect-Based Sentiment Analysis

2026-07-15

AI Summary

Existing counterfactual generation methods struggle to flip the sentiment of a target aspect while simultaneously preserving sentiments of other aspects, maintaining semantic consistency, fluency, and factual correctness, thereby limiting their effectiveness in evaluating the robustness of aspect-based sentiment analysis (ABSA) models. To address this, this work proposes CAVE-ABSA, a novel framework that decouples counterfactual generation and verification for the first time. It localizes opinion spans corresponding to the target aspect, performs controlled rewriting, and incorporates a repair module. Additionally, it introduces a multidimensional verification mechanism that integrates constraints such as non-target aspect sentiment preservation, AMR structural consistency, minimal edit distance, and contradiction detection. This approach yields a high-quality, task-aligned aspect-level counterfactual dataset, significantly enhancing the reliability of robustness and reasoning evaluations for ABSA models.

Abstract

Aspect-Based Sentiment Analysis (ABSA) requires models to identify sentiment toward specific aspects rather than relying on the global polarity of a sentence. This makes counterfactual evaluation especially challenging: a valid counterfactual should flip the sentiment of one target aspect while preserving the sentiment of all non-target aspects, semantic meaning, fluency, and factual consistency. Existing counterfactual generation methods often focus on sentence-level label flipping and may produce edits that are fluent but aspect-invalid, semantically drifting, or contradictory. To address this limitation, we propose CAVE-ABSA, a Constraint-Aware Validated Editing framework for generating and validating aspect-level counterfactuals. CAVE-ABSA localizes the opinion span associated with the target aspect, performs controlled counterfactual rewriting, refines candidates through a repair module, and filters them using aspect-level verification, semantic similarity, AMR-guided structural preservation, edit minimality, fluency, and contradiction detection. The framework is designed to construct validated counterfactual ABSA datasets for robustness evaluation and data augmentation. By explicitly separating generation from validation, CAVE-ABSA provides a principled approach for producing meaningful aspect-local counterfactuals and for testing whether ABSA models truly rely on aspect-grounded sentiment reasoning.

Problem

• Aspect-Based Sentiment Analysis
• Counterfactual Generation
• Sentiment Consistency
• Semantic Preservation
• Factual Consistency

Innovation

• Counterfactual Editing
• Aspect-Based Sentiment Analysis
• Constraint-Aware Generation
• AMR-Guided Validation
• Opinion Span Localization