NCL-BU at SemEval-2026 Task 3: Fine-tuning XLM-RoBERTa for Multilingual Dimensional Sentiment Regression

2026-04-10

AI Summary

This study addresses multilingual dimensional aspect-based sentiment analysis, aiming to predict real-valued scores for each aspect in a text along two continuous affective dimensions—valence and arousal—within the range [1, 9]. Building upon XLM-RoBERTa-base, the authors employ a task-specific fine-tuning strategy that structures inputs as [CLS] text [SEP] aspect [SEP] and introduces a dual regression head with Sigmoid scaling to map outputs to the target range. Separate models are trained for English and Chinese across three domains: restaurants, laptops, and finance, with training and development sets merged for final evaluation. The work provides the first systematic demonstration that fine-tuning consistently and significantly outperforms few-shot prompting with prominent large language models such as GPT-5.2 and LLaMA-3/4, thereby establishing the superiority of fine-tuned approaches in multilingual dimensional sentiment regression.

Technology Category

• Natural Language Processing: Sentiment Analysis, Stylistic Analysis, and Argument Mining
• Machine Learning: Large Multimodal Models (LMMs)
• Cognitive Modeling & Cognitive Systems: Affective Computing

Application Category

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

Abstract

Dimensional Aspect-Based Sentiment Analysis (DimABSA) extends traditional ABSA from categorical polarity labels to continuous valence-arousal (VA) regression. This paper describes a system developed for Track A - Subtask 1 (Dimensional Aspect Sentiment Regression), aiming to predict real-valued VA scores in the [1, 9] range for each given aspect in a text. A fine-tuning approach based on XLM-RoBERTa-base is adopted, constructing the input as [CLS] T [SEP] a_i [SEP] and training dual regression heads with sigmoid-scaled outputs for valence and arousal prediction. Separate models are trained for each language-domain combination (English and Chinese across restaurant, laptop, and finance domains), and training and development sets are merged for final test predictions. In development experiments, the fine-tuning approach is compared against several large language models including GPT-5.2, LLaMA-3-70B, LLaMA-3.3-70B, and LLaMA-4-Maverick under a few-shot prompting setting, demonstrating that task-specific fine-tuning substantially and consistently outperforms these LLM-based methods across all evaluation datasets. The code is publicly available at https://github.com/tongwu17/SemEval-2026-Task3-Track-A.

Problem

• Dimensional Aspect-Based Sentiment Analysis
• Valence-Arousal Regression
• Multilingual Sentiment Analysis
• Aspect-Level Sentiment

Innovation

• Dimensional Aspect-Based Sentiment Analysis
• XLM-RoBERTa fine-tuning
• Valence-Arousal regression
• Multilingual sentiment analysis
• Task-specific fine-tuning