Lotus at SemEval-2025 Task 11: RoBERTa with Llama-3 Generated Explanations for Multi-Label Emotion Classification

TL;DR

This paper introduces a novel multi-label emotion classification method that uses Llama-3 generated explanations to improve RoBERTa's performance, especially on ambiguous emotional expressions, achieving higher F1-scores.

Contribution

It presents a new approach combining Llama-3 explanations with RoBERTa for enhanced emotion detection, addressing ambiguity and overlapping cues in multi-label classification.

Findings

Improved F1-scores for fear, joy, sadness

Outperforms text-only models

Enhances multi-label emotion classification

Abstract

This paper presents a novel approach for multi-label emotion detection, where Llama-3 is used to generate explanatory content that clarifies ambiguous emotional expressions, thereby enhancing RoBERTa's emotion classification performance. By incorporating explanatory context, our method improves F1-scores, particularly for emotions like fear, joy, and sadness, and outperforms text-only models. The addition of explanatory content helps resolve ambiguity, addresses challenges like overlapping emotional cues, and enhances multi-label classification, marking a significant advancement in emotion detection tasks.