YangSteam at SemEval-2026 Task 3: Transformer-Based Aspect-Aware Regression for Dimensional Sentiment and Stance Analysis

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, quick

Summary

YangSteam's system for SemEval-2026 Task 3 addresses Dimensional Aspect-Based Sentiment Analysis (DimABSA) and Dimensional Stance Analysis (DimStance). The system predicts continuous valence–arousal (VA) scores for text–aspect pairs in English and Chinese, participating in Track A and Track B. It integrates pre-trained multilingual transformers with aspect-marker input encoding and dual regression heads, trained via a 5-fold cross-validation ensemble. XLM-RoBERTa-large serves as the backbone for Track A, while mDeBERTa-v3-base is used for Track B, selected after systematic model comparison. The system significantly outperformed organizer-provided baselines on official test sets and achieved strong results on the unofficial post-evaluation leaderboard, ranking 1st on zho-env (Track B) and 2nd on zho-fin (Track A).

Key takeaway

For NLP Engineers developing dimensional sentiment or stance analysis systems, especially in multilingual contexts, consider integrating aspect-marker input encoding with pre-trained transformers and dual regression heads. Your model selection should involve systematic comparison, as demonstrated by the use of XLM-RoBERTa-large for DimABSA and mDeBERTa-v3-base for DimStance. Employing ensemble methods like 5-fold cross-validation can significantly improve performance and robustness, particularly for challenging language subsets.

Key insights

Multilingual transformers combined with aspect-marker encoding and dual regression effectively predict dimensional sentiment and stance.

Principles

• Systematic model comparison optimizes backbone selection.
• Ensemble methods enhance prediction robustness.
• Aspect-marker encoding improves contextual understanding.

Method

The system uses pre-trained multilingual transformers, aspect-marker input encoding, and dual regression heads for VA prediction, trained with a 5-fold cross-validation ensemble.

In practice

• Consider XLM-RoBERTa-large for DimABSA tasks.
• Evaluate mDeBERTa-v3-base for DimStance tasks.
• Apply 5-fold cross-validation for robust training.

Topics

• Dimensional Sentiment Analysis
• Stance Analysis
• Multilingual Transformers
• XLM-RoBERTa-large
• mDeBERTa-v3-base
• Ensemble Learning

Best for: Research Scientist, AI Scientist, NLP Engineer, Machine Learning Engineer

Related on AIssential

• From TF-IDF to Transformers: A Comparative and Ensemble Approach to Sentiment Classification — A soft voting ensemble of diverse machine learning and transformer models significantly enhances sentiment classification accuracy, particularly for nuanced text.
• Fine-Tuning BERT for Sentiment Analysis: A Beginner-Friendly Guide — Fine-tuning pre-trained transformer models like BERT enables high-accuracy NLP tasks with minimal training data.
• Teaching Machines to “Feel” Tweets — Domain-native pre-trained models are crucial for accurate sentiment classification in ambiguous social media language.
• Zero-Shot to Full-Resource: Cross-lingual Transfer Strategies for Aspect-Based Sentiment Analysis — Cross-lingual transfer strategies for ABSA vary in effectiveness based on model architecture and resource availability.
• Leveraging political alignment information for stance detection — Political alignment can infer stance on related issues, reducing the need for new labelled data.
• Sentiment Analysis with BERT: The Heart of the Transformer, The Soul of the Text — BERT's bidirectional context understanding via Transformers significantly advanced sentiment analysis and NLP tasks.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.