Can We Identify Stance Without Target Arguments? A Study for Rumour Stance Classification
Yue Li, Carolina Scarton
TL;DR
The paper investigates rumour stance classification in Twitter threads, revealing that many replies can indicate stance without the rumour target and that target-aware models can underperform when the target context is essential. It analyzes target-dependent vs target-independent cues via annotation and comprehensive model evaluations, showing strong performance by target-oblivious approaches on target-independent data. To address the limitations of existing methods, it introduces an ensemble framework that combines a cross-attention Siamese encoder of source and reply with a sample-reweighting strategy to emphasize target-dependent cases, achieving state-of-the-art results on RumourEval 2017 and 2019. The work provides practical guidance for leveraging target information in rumor verification and releases annotation data to foster further research.
Abstract
Considering a conversation thread, rumour stance classification aims to identify the opinion (e.g. agree or disagree) of replies towards a target (rumour story). Although the target is expected to be an essential component in traditional stance classification, we show that rumour stance classification datasets contain a considerable amount of real-world data whose stance could be naturally inferred directly from the replies, contributing to the strong performance of the supervised models without awareness of the target. We find that current target-aware models underperform in cases where the context of the target is crucial. Finally, we propose a simple yet effective framework to enhance reasoning with the targets, achieving state-of-the-art performance on two benchmark datasets.