MSMO-ABSA: Multi-Scale and Multi-Objective Optimization for Cross-Lingual Aspect-Based Sentiment Analysis

TL;DR

This paper introduces MSMO, a novel framework for cross-lingual aspect-based sentiment analysis that employs multi-scale feature alignment and multi-objective optimization to improve performance across languages.

Contribution

The paper proposes a new MSMO framework that combines multi-scale alignment, bilingual code-switching, and knowledge distillation for enhanced cross-lingual ABSA.

Findings

MSMO achieves state-of-the-art results across multiple languages.

Incorporating code-switched sentences improves robustness.

Multi-objective optimization enhances semantic alignment.

Abstract

Aspect-based sentiment analysis (ABSA) garnered growing research interest in multilingual contexts in the past. However, the majority of the studies lack more robust feature alignment and finer aspect-level alignment. In this paper, we propose a novel framework, MSMO: Multi-Scale and Multi-Objective optimization for cross-lingual ABSA. During multi-scale alignment, we achieve cross-lingual sentence-level and aspect-level alignment, aligning features of aspect terms in different contextual environments. Specifically, we introduce code-switched bilingual sentences into the language discriminator and consistency training modules to enhance the model's robustness. During multi-objective optimization, we design two optimization objectives: supervised training and consistency training, aiming to enhance cross-lingual semantic alignment. To further improve model performance, we incorporate distilled knowledge of the target language into the model. Results show that MSMO significantly enhances cross-lingual ABSA by achieving state-of-the-art performance across multiple languages and models.