Zero-Shot Stance Detection: A Dataset and Model using Generalized Topic Representations

Stance detection is an important component of understanding hidden influences\nin everyday life. Since there are thousands of potential topics to take a\nstance on, most with little to no training data, we focus on zero-shot stance\ndetection: classifying stance from no training examples. In this paper, we\npresent a new dataset for zero-shot stance detection that captures a wider\nrange of topics and lexical variation than in previous datasets. Additionally,\nwe propose a new model for stance detection that implicitly captures\nrelationships between topics using generalized topic representations and show\nthat this model improves performance on a number of challenging linguistic\nphenomena.\n