HARGAN: Heterogeneous Argument Attention Network for Persuasiveness Prediction

Abstract

Argument structure elaborates the relation among claims and premises. Previous works in persuasiveness prediction do not consider this relation in their architectures. To take argument structure information into account, this paper proposes an approach to persuasiveness prediction with a novel graph-based neural network model, called heterogeneous argument attention network (HARGAN). By jointly training on the persuasiveness and stance of the replies, our model achieves the state-of-the-art performance on the ChangeMyView (CMV) dataset for the persuasiveness prediction task. Experimental results show that the graph setting enables our model to aggregate information across multiple paragraphs effectively. In the meanwhile, our stance prediction auxiliary task enables our model to identify the viewpoint of each party, and helps our model perform better on the persuasiveness prediction.

Heterogeneous Graph Contrastive Learning for Stance Prediction

“…For stance prediction, most studies mainly predict the stances of users from micro and macro levels in social media platforms [4], [9] or online debate forums [8], [10]. From the microscopic perspective, stance prediction can be viewed as a recommendation problem.…”

“…By analyzing the experimental results, it can be seen that the historical behavior of users is helpful to predict the stance of "silent users". Huang et al [10] proposed a Heterogeneous Argument Attention Network, which jointly learned stance prediction and persuasiveness prediction in muti-round of dialogues. The advantage of this model lies in that it makes use of argument structure information through the GNN module.…”

“…Instead, they may show endorsements of their support or opposition to the current issue by agreeing with some other issues or other users. Therefore, inferring the stance of "silent users" has attracted much attention in recent years [8]- [10] and the task of stance prediction has emerged.…”