Home // eKNOW 2025, The Seventeenth International Conference on Information, Process, and Knowledge Management // View article

Stance-Conditioned Modeling for Rumor Verification

Authors:
Gibson Nkhata
Susan Gauch

Keywords: Rumor verification; stance-conditioned modeling; social media misinformation; embedding aggregation.

Abstract:
The rapid spread of misinformation on social media platforms has heightened the need for effective rumor verification models. Traditional approaches primarily rely on textual content and transformer-based embeddings, but they often fail to incorporate conversational dynamics and stance evolution, limiting their effectiveness. We present a stance-conditioned rumor verification model that integrates Bidirectional Encoder Representations from Transformers (BERT) based source post embeddings, reply post embedding aggregation, and Bidirectional Long Short Term Memory (BiLSTM) encoding of stance labels to enhance rumor classification. By explicitly modeling stance progression and leveraging aggregated stance-conditioned reply embeddings, our approach captures critical discourse patterns that influence rumor veracity. Experiments on competitive benchmark tasks demonstrate that our model outperforms state-of-the-art baselines in Macro-F1 and accuracy, achieving superior performance across multiple datasets. Ablation studies confirm the effectiveness of each constituent model component, with early rumor detection analysis showcasing our model’s ability to detect misinformation faster and more accurately than competing methods. Overall, this work presents a novel stance-conditioned approach to rumor verification that effectively captures conversational context and discourse interactions, providing a more robust and interpretable framework for combating online misinformation.

Pages: 30 to 36

Copyright: Copyright (c) IARIA, 2025

Publication date: May 18, 2025

Published in: conference

ISSN: 2308-4375

ISBN: 978-1-68558-272-2

Location: Nice, France

Dates: from May 18, 2025 to May 22, 2025