Abstract: Stance detection is an important task, supporting many downstream tasks such as discourse parsing and modeling the propagation of fake news, rumors, and science denial. In this talk we describe a novel framework for stance detection. Our framework is unsupervised and domain-independent. Given a claim and a multi-participant discussion – we construct the interaction network from which we derive topological embeddings for each speaker. These speaker embeddings enjoy the following property: speakers with the same stance tend to be represented by similar vectors, while antipodal vectors represent speakers with opposing stances. These embeddings are then used to divide the speakers into stance-partitions. Our embedding is derived from the Semi-Definite Programming (SDP) solution to the max-cut problem on the interaction network. In this talk we shall explain how the success of our method is rooted in theoretical results of SDP integrality for random k-colorable graphs.
We evaluated our method on three different datasets from different platforms. Our method outperforms or is comparable with supervised models, including Neural Network based approaches.