Disentangled VAD Representations via a Variational Framework for Political Stance Detection

TL;DR

This paper tackles stance detection in political discourse by integrating granular sentiment with latent affective representations. The authors introduce PoliStance-VAE, a variational autoencoder that disentangles Valence, Arousal, and Dominance (VAD) while reconstructing text, aided by an auxiliary sentiment prediction trained using seven-class annotations from EmoLLMs. Through a target-aware encoder and a VAD-supervised learning objective, the model achieves state-of-the-art results on P-STANCE and SemEval-2016, and ablation confirms the importance of VAD and sentiment supervision for performance and generalization. The approach demonstrates the value of fine-grained emotional representations for cross-target stance detection and lays groundwork for broader NLP tasks requiring nuanced affective understanding.

Abstract

The stance detection task aims to categorise the stance regarding specified targets. Current methods face challenges in effectively integrating sentiment information for stance detection. Moreover, the role of highly granular sentiment labelling in stance detection has been largely overlooked. This study presents a novel stance detection framework utilizing a variational autoencoder (VAE) to disentangle latent emotional features-value, arousal, and dominance (VAD)-from political discourse on social media. This approach addresses limitations in current methods, particularly in in-target and cross-target stance detection scenarios. This research uses an advanced emotional annotation tool to annotate seven-class sentiment labels for P-STANCE. Evaluations on benchmark datasets, including P-STANCE and SemEval-2016, reveal that PoliStance-VAE achieves state-of-the-art performance, surpassing models like BERT, BERTweet, and GPT-4o. PoliStance-VAE offers a robust and interpretable solution for stance detection, demonstrating the effectiveness of integrating nuanced emotional representations. This framework paves the way for advancements in natural language processing tasks, particularly those requiring detailed emotional understanding.

Figures (4)

• Figure 1: Example of Data
• Figure 2: Main components of PoliStance-VAE.
• Figure 3: Sentiment distribution across the targets in P-STANCE.
• Figure 4: Sentiment distribution across the targets in SemEval-2016.