A Node Classification Approach for Dynamically Extracting the Structures of Online Discussions

Abstract

Online discussion platforms require extracting the discussion structure in order to support understanding the flow of these discussions. Towards this end, this paper proposes an approach that performs node classification which is the first step towards extracting the structure of online discussions. In this regard, the proposed approach employs a graph attention network (GAT) in order to directly learn the discussion structure. In specific, the GAT, which is a type of graph neural networks (GNNs), encodes the graph structures directly. In addition, the GAT, which is based on attention architecture, is able to deal with different graph structures. In order to evaluate the proposed approach, we have conducted a set of experiments on the persuasive essays dataset that is styled using the issue-based information system (IBIS). The experimental results show that the proposed approach is able to classify the nodes in online discussion structures accurately.