NCL-BU at SemEval-2026 Task 3: Fine-tuning XLM-RoBERTa for Multilingual Dimensional Sentiment Regression
2026-04-10 • Computation and Language
Computation and Language
AI summaryⓘ
The authors worked on a system to predict feelings about specific parts of a text, not just as positive or negative, but with detailed scores for how good or exciting those parts are. They used a special language model called XLM-RoBERTa and taught it to give two numbers for each part of the text, one for 'valence' (how positive) and one for 'arousal' (how intense). They trained separate models for English and Chinese in different topics like restaurants and finance. Their method did better than using very large language models without extra training on this specific task. They also shared their code publicly.
Dimensional Aspect-Based Sentiment AnalysisValence-ArousalXLM-RoBERTaFine-tuningSentiment RegressionAspect-Based Sentiment AnalysisFew-shot promptingMultilingual modelsNatural Language ProcessingSemEval
Authors
Tong Wu, Nicolay Rusnachenko, Huizhi Liang
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.