Adversarial Learning for Zero-Shot Stance Detection on Social Media

Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 4756–4767 June 6–11, 2021. ©2021 Association for Computational Linguistics 4756 Adversarial Learning for Zero-Shot Stance Detection on Social Media Emily Allaway⇤ Malavika Srikanth⇤ Department of Computer Science, Columbia University, New York, NY {eallaway,kathy}@cs.columbia.edu, ms5908@columbia.edu Kathleen McKeown Abstract Stance detection on social media can help to identify and understand slanted news or com- mentary in everyday life. In this work, we pro- pose a new model for zero-shot stance detec- tion on Twitter that uses adversarial learning to generalize across topics. Our model achieves state-of-the-art performance on a number of unseen test topics with minimal computational costs. In addition, we extend zero-shot stance detection to new topics, highlighting future di- rections for zero-shot transfer. 1 Introduction Stance detection, the problem of automatically identifying positions or opinions in text, is becom- ing increasingly important for social media (e.g., Twitter), as more and more people turn to it for their news. Zero-shot stance detection, in particular, is crucial, since gathering training data for all topics is not feasible. While there has been increasing work on zero-shot stance detection in other gen- res (Allaway and McKeown, 2020; Vamvas and Sennrich, 2020), generalization across many topics in social media remains an open challenge. In this work, we propose a new model for stance detection that uses adversarial learning to general- ize to unseen topics on Twitter. Our model achieves state-of-the-art zero-shot performance on the ma- jority of topics in the standard dataset for English stance detection on Twitter (Mohammad et al., 2016) and also provides benchmark results on two new topics in this dataset. Most prior work on English social media stance detection uses the SemEval2016 Task 6 (SemT6) dataset (Mohammad et al., 2016) which consists of six topics. While early work trained using five topics and evaluated on the sixth (e.g., Augenstein et al. (2016); Zarrella and Marsh (2016); Wei et al. (2016)), they used only one topic, ‘Donald Trump’ ⇤⇤ Denotes equal contribution. (DT), for evaluation and did not experiment with others. Furthermore, recent work on SemT6 has focused on cross-target stance detection (Xu et al., 2018; Wei and Mao, 2019; Zhang et al., 2020): training on one topic and evaluating on one dif- ferent unseeen topic that has a known relation- ship with the training topic (e.g., “legalization of abortion” to “feminist movement”). These models are typically evaluated on four different test topics (each with a different training topic). In contrast, our work is a hybrid of these two set- tings: we train on five topics and evaluate on one other, but unlike prior work we do not assume a relationship between training and test topics and so we use each topic in turn as the test topic. This illus- trates the robustness of our model across topics and additionally allows zero-shot evaluation of SemT6 on two new topics that were previously ignored by cross-target models (‘atheism’ and ‘climate change is a …

Content truncated to 500 words