SCUZANE at SemEval-2026 Task 3: Dimensional Aspect-based Sentiment Analysis with Supervised Contrastive Regression and R-Drop Regularization

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

Summary

The SCUZANE system, presented at SemEval-2026 Task 3, addresses the limitations of traditional Aspect-Based Sentiment Analysis (ABSA) by focusing on Dimensional ABSA (DimABSA). This approach replaces coarse-grained categorical sentiment labels like "Positive" or "Negative" with continuous valence-arousal (VA) scores to capture subtle emotional intensity in text. Specifically designed for Subtask 1 (Dimensional Aspect Sentiment Regression) of Track A, SCUZANE utilizes a DeBERTa-v3-large backbone. Its architecture is enhanced by a prompt-based learning strategy that concatenates aspect information with the context. Furthermore, the system incorporates multi-sample dropout and a weighted aggregation of hidden states from the last four layers to derive rich semantic representations. Experimental results across various domains and languages confirm the effectiveness of integrating consistency regularization with dimensional contrastive learning for fine-grained sentiment regression.

Key takeaway

For NLP engineers developing fine-grained sentiment analysis systems, consider adopting dimensional aspect-based sentiment regression. Your current categorical models may miss subtle emotional nuances; shifting to continuous valence-arousal scores can provide richer insights. Implement prompt-based learning with a robust backbone like DeBERTa-v3-large, and integrate consistency regularization with dimensional contrastive learning to enhance model performance and capture deeper emotional intensity.

Key insights

Dimensional ABSA with contrastive regression and R-Drop regularization effectively captures fine-grained emotional intensity using continuous VA scores.

Principles

• Continuous VA scores enhance sentiment granularity.
• Prompt-based learning improves aspect-context integration.
• Consistency regularization boosts regression performance.

Method

The SCUZANE system uses a DeBERTa-v3-large backbone, prompt-based learning to concatenate aspect info with context, multi-sample dropout, and weighted aggregation of the last four hidden layers for semantic representation. It integrates consistency regularization with dimensional contrastive learning.

In practice

• Implement prompt-based learning for aspect integration.
• Apply multi-sample dropout for robust representations.
• Use weighted aggregation of transformer hidden states.

Topics

• Dimensional ABSA
• Sentiment Analysis
• Supervised Contrastive Regression
• R-Drop Regularization
• DeBERTa-v3-large
• Prompt-based Learning

Best for: Research Scientist, AI Scientist, NLP Engineer

Related on AIssential

• DiffuSent: Towards a Unified Diffusion Framework for Aspect-Based Sentiment Analysis — DiffuSent unifies ABSA subtasks via non-auto-regressive boundary denoising, improving accuracy and speed for multi-word sentiment extraction.
• Beyond Sentiment Classification: A Generative Framework for Emotion Intensity Evaluation in Text — Generative LLMs fine-tuned on continuous emotion intensity scores outperform classification and generalize to unseen affective dimensions.
• Aspect-Based Sentiment Analysis in the Accommodation Domain Using the BERTimbau Model — BERTimbau model fine-tuning strategies for Portuguese ABSA show trade-offs between recall and precision.
• Consistency Analysis of Sentiment Predictions using Syntactic & Semantic Context Assessment Summarization (SSAS) — SSAS improves LLM sentiment prediction consistency by pre-processing noisy data into structured, high-signal prompts.
• Learning Word Vectors for Sentiment Analysis: A Python Reproduction — Word vectors can effectively capture both semantic similarity and sentiment polarity by combining unsupervised and supervised learning objectives.
• Turkish Sentiment Analysis (Keras LSTM + BERT) — A hybrid NLP pipeline combines custom LSTMs and pre-trained BERT for robust Turkish sentiment analysis with real-time data acquisition.

Open in AIssential →

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