AILS-NTUA at SemEval-2026 Task 3: Efficient Dimensional Aspect-Based Sentiment Analysis
arXiv:2603.04933v1 Announce Type: new Abstract: In this paper, we present AILS-NTUA system for Track-A of SemEval-2026 Task 3 on Dimensional Aspect-Based Sentiment Analysis (DimABSA), which encompasses three complementary problems: Dimensional Aspect Sentiment Regression (DimASR), Dimensional Aspect Sentiment Triplet Extraction (DimASTE), and Dimensional Aspect Sentiment Quadruplet Prediction (DimASQP) within a multilingual and multi-domain framework. Our methodology combines fine-tuning of language-appropriate encoder backbones for continuous aspect-level sentiment prediction with language-specific instruction tuning of large language models using LoRA for structured triplet and quadruplet extraction. This unified yet task-adaptive design emphasizes parameter-efficient specialization across languages and domains, enabling reduced training and inference requirements while maintaining strong effectiveness. Empirical results demonstrate that the proposed models achieve competitive perfo
arXiv:2603.04933v1 Announce Type: new Abstract: In this paper, we present AILS-NTUA system for Track-A of SemEval-2026 Task 3 on Dimensional Aspect-Based Sentiment Analysis (DimABSA), which encompasses three complementary problems: Dimensional Aspect Sentiment Regression (DimASR), Dimensional Aspect Sentiment Triplet Extraction (DimASTE), and Dimensional Aspect Sentiment Quadruplet Prediction (DimASQP) within a multilingual and multi-domain framework. Our methodology combines fine-tuning of language-appropriate encoder backbones for continuous aspect-level sentiment prediction with language-specific instruction tuning of large language models using LoRA for structured triplet and quadruplet extraction. This unified yet task-adaptive design emphasizes parameter-efficient specialization across languages and domains, enabling reduced training and inference requirements while maintaining strong effectiveness. Empirical results demonstrate that the proposed models achieve competitive performance and consistently surpass the provided baselines across most evaluation settings.
Executive Summary
The AILS-NTUA system tackles Dimensional Aspect-Based Sentiment Analysis, a multilingual and multi-domain task, by fine-tuning encoder backbones and utilizing large language models. This approach enables parameter-efficient specialization, reducing training and inference requirements while maintaining strong performance. The system achieves competitive results, surpassing baselines in most evaluation settings.
Key Points
- ▸ Combination of fine-tuning and language-specific instruction tuning
- ▸ Use of LoRA for structured triplet and quadruplet extraction
- ▸ Unified yet task-adaptive design for multilingual and multi-domain framework
Merits
Parameter Efficiency
The system's design allows for reduced training and inference requirements, making it more efficient and scalable.
Competitive Performance
The proposed models achieve competitive results, consistently surpassing the provided baselines across most evaluation settings.
Demerits
Complexity
The system's reliance on fine-tuning and large language models may introduce complexity, potentially limiting its accessibility and interpretability.
Expert Commentary
The AILS-NTUA system demonstrates a nuanced understanding of the complexities involved in dimensional aspect-based sentiment analysis. By leveraging fine-tuning and language-specific instruction tuning, the system achieves a balance between parameter efficiency and competitive performance. However, the system's reliance on large language models and fine-tuning may introduce complexity, highlighting the need for further research into interpretability and accessibility. Overall, the system's design and results contribute significantly to the field of natural language processing, with potential applications in sentiment analysis and opinion mining.
Recommendations
- ✓ Further investigation into the interpretability and accessibility of the system's design
- ✓ Exploration of the system's potential applications in social media monitoring and customer feedback analysis