Khaleesiyali at SemEval-2026 Task 2: Lexicon-Augmented RoBERTa for Valence–Arousal Regression on Ecological Essays

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

Summary

The Khaleesiyali system, presented at SemEval-2026 Task 2, addresses the valence–arousal regression challenge on ecological essays. This system integrates a RoBERTa model with a 6-dimensional feature vector derived from the NRC VAD lexicon, achieving a high token coverage rate of 72.05%. Under official user-aware evaluation, the lexicon-augmented RoBERTa system achieved a competitive average composite correlation of 0.547. It significantly outperformed the ridge regression baseline, demonstrating particular robustness in valence (r = 0.656) and strong generalization to unseen users (rarousal = 0.519). These findings suggest that lightweight lexicon-based statistics offer valuable complementary cues for longitudinal emotion modeling within modern transformer architectures.

Key takeaway

For NLP engineers developing emotion recognition systems, consider augmenting transformer models like RoBERTa with lightweight lexicon-based features. This approach, demonstrated by the Khaleesiyali system's 0.547 composite correlation, significantly improves valence robustness (r=0.656) and generalization to unseen users (rarousal=0.519). Your models will benefit from these complementary cues, potentially reducing the need for extensive task-specific fine-tuning and enhancing performance on longitudinal emotion modeling.

Key insights

Lexicon-augmented RoBERTa models enhance valence-arousal regression by integrating lightweight lexicon-based features, improving generalization.

Principles

• Lexicon-based statistics complement transformer embeddings.
• Augmenting deep models improves emotion modeling robustness.
• Generalization to unseen users is achievable with lexicon features.

Method

Integrate a 6-dimensional NRC VAD lexicon feature vector with deep contextual embeddings from a RoBERTa model for valence–arousal regression.

In practice

• Combine RoBERTa with NRC VAD lexicon features.
• Apply lexicon augmentation for emotion regression tasks.
• Evaluate models on user-aware metrics for robustness.

Topics

• SemEval-2026 Task 2
• Valence–Arousal Regression
• RoBERTa
• NRC VAD Lexicon
• Emotion Modeling
• Transformer Architectures
• Lexicon Augmentation

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

Related on AIssential

• 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.
• Exploration of symbolic methods for emotion detection for Portuguese — Symbolic emotion detection in Portuguese shows stable relative performance despite domain-specific absolute variations.
• Emotion-Cause Pair Extraction in Conversations via Semantic Decoupling and Graph Alignment — Decoupling emotion and cause semantics improves conversational emotion-cause pair extraction.
• 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.
• From Affect Prediction to Affect Forecasting: Evidence for Distinct Information Sources in Longitudinal Text — Current affect prediction relies on text semantics, while future affective change forecasting benefits from numeric trajectory dynamics.
• A Scalable Tool for Measuring Manner and Result Verbs in Developmental Language Research — A RoBERTa classifier, trained on LLM-generated annotations, accurately distinguishes manner and result verbs at scale.

Open in AIssential →

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