ICT-NLP at SemEval-2026 Task 3: Less Is More -- Multilingual Encoder with Joint Training and Adaptive Ensemble for Dimensional Aspect Sentiment Regression

2026-05-11

AI Summary

This work addresses the challenges of data sparsity and cross-lingual transfer in dimensional aspect-based sentiment regression across multiple languages and domains. The authors propose a lightweight solution that does not rely on large language models or external corpora. Building upon a multilingual pretrained encoder, the approach integrates joint multilingual and multidomain training, employs a bounded regression transformation to enhance training stability, and introduces an adaptive ensemble strategy based on subset search to reduce prediction variance. Evaluated on SemEval-2026 Task 3, the method achieves top performance: first place on the Chinese restaurant dataset, second on the Chinese laptop dataset, third on the Japanese hotel dataset, and ranks within the top 50% of all participating systems on all remaining datasets.

Abstract

This paper describes our system to SemEval-2026 Task 3 Track A Subtask 1 on Dimensional Aspect Sentiment Regression (DimASR). We propose a lightweight and resource-efficient system built entirely on multilingual pre-trained encoders, without relying on LLMs or external corpora. We adopt joint multilingual and multi-domain training to facilitate cross-lingual transfer and alleviate data sparsity, introduce a bounded regression transformation that improves training stability while constraining predictions within the valid range, and employ an adaptive ensemble strategy via subset search to reduce prediction variance. Experimental results demonstrate that our system achieves strong and consistent performance, ranking 1st on zho-res, 2nd on zho-lap, and 3rd on jpn-hot, with all remaining datasets placed within the top half of participating teams.

Problem

• Dimensional Aspect Sentiment Regression
• Multilingual
• Cross-lingual Transfer
• Data Sparsity
• Sentiment Analysis

Innovation

• multilingual encoder
• joint training
• bounded regression
• adaptive ensemble
• dimensional sentiment regression