Zero-Shot to Full-Resource: Cross-lingual Transfer Strategies for Aspect-Based Sentiment Analysis

2026-04-29

AI Summary

This study addresses the scarcity of non-English resources and English-centric bias in multilingual fine-grained aspect-based sentiment analysis (ABSA) by systematically evaluating cross-lingual transfer strategies across seven languages and four subtasks under zero-resource, target-language-only, and full-resource settings. Through comprehensive comparisons of approaches—including fine-tuning large language models (LLMs), few-shot prompting, multilingual joint training, code-switching, and machine translation—the work finds that LLM fine-tuning achieves the best performance on complex generative tasks, while few-shot methods nearly match its efficacy on simpler tasks. Multilingual joint training proves most compatible with LLMs, whereas code-switching benefits smaller encoders or seq-to-seq architectures more effectively. To advance non-English ABSA research, the paper introduces the first German ASQP dataset, GERest, along with an enhanced version, GERestaurant.

Abstract

Aspect-based Sentiment Analysis (ABSA) extracts fine-grained opinions toward specific aspects within text but remains largely English-focused despite major advances in transformer-based and instruction-tuned models. This work presents a multilingual evaluation of state-of-the-art ABSA approaches across seven languages (English, German, French, Dutch, Russian, Spanish, and Czech) and four subtasks (ACD, ACSA, TASD, ASQP). We systematically compare different transformer architectures under zero-resource, data-only, and full-resource settings, using cross-lingual transfer, code-switching and machine translation. Fine-tuned Large Language Models (LLMs) achieve the highest overall scores, particularly in complex generative tasks, while few-shot counterparts approach this performance in simpler setups, where smaller encoder models also remain competitive. Cross-lingual training on multiple non-target languages yields the strongest transfer for fine-tuned LLMs, while smaller encoder or seq-to-seq models benefit most from code-switching, highlighting architecture-specific strategies for multilingual ABSA. We further contribute two new German datasets, an adapted GERestaurant and the first German ASQP dataset (GERest), to encourage multilingual ABSA research beyond English.

Problem

• Aspect-Based Sentiment Analysis
• Cross-lingual Transfer
• Multilingual NLP
• Zero-Shot Learning
• Full-Resource Setting

Innovation

• cross-lingual transfer
• aspect-based sentiment analysis
• large language models
• code-switching
• multilingual evaluation