AILS-NTUA at SemEval-2026 Task 3: Efficient Dimensional Aspect-Based Sentiment Analysis

2026-03-05

AI Summary

This work addresses dimensional aspect-based sentiment analysis in multilingual and multidomain settings by unifying three subtasks: sentiment regression, triplet extraction, and quadruplet prediction. The authors propose a parameter-efficient joint framework that combines language-adaptive encoder fine-tuning to support continuous sentiment prediction with instruction tuning of large language models via LoRA for effective extraction of structured sentiment triplets and quadruplets. The approach achieves competitive performance while significantly reducing training and inference costs, consistently outperforming existing baselines across most evaluation settings. It demonstrates strong generalization capabilities across languages and domains, highlighting its effectiveness and efficiency in complex, real-world sentiment analysis scenarios.

Technology Category

• Natural Language Processing: Sentiment Analysis, Stylistic Analysis, and Argument Mining
• Machine Learning: Large Multimodal Models (LMMs)
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data

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.

Problem

• Dimensional Aspect-Based Sentiment Analysis
• Aspect Sentiment Regression
• Aspect Sentiment Triplet Extraction
• Aspect Sentiment Quadruplet Prediction
• Multilingual Multi-domain

Innovation

• Dimensional Aspect-Based Sentiment Analysis
• LoRA
• Parameter-Efficient Fine-Tuning
• Multilingual Multi-Domain
• Structured Sentiment Extraction