The ongoing COVID-19 pandemic has caused immeasurable losses for people worldwide. To contain the spread of virus and further alleviate the crisis, various health policies (e.g., stay-at-home orders) have been issued which spark heat discussion as users turn to share their attitudes on social media. In this paper, we consider a more realistic scenario on stance detection (i.e., cross-target and zero-shot settings) for the pandemic and propose an adversarial learning-based stance classifier to automatically identify the public attitudes toward COVID-19-related health policies. Specifically, we adopt adversarial learning which allows the model to train on a large amount of labeled data and capture transferable knowledge from source topics, so as to enable generalize to the emerging health policy with sparse labeled data. Meanwhile, a GeoEncoder is designed which encourages model to learn unobserved contextual factors specified by each region and represents them as non-text information to enhance model's deeper understanding. We evaluate the performance of a broad range of baselines in stance detection task for COVID-19-related policies, and experimental results show that our proposed method achieves state-of-the-art performance in both cross-target and zero-shot settings.
翻译:为遏制病毒的传播并进一步缓解危机,发布了各种卫生政策(如在家停留令),在用户转向社交媒体交流其态度时引发热讨论。在本文件中,我们认为该流行病的立体检测(即交叉目标和零射环境)的情景更为现实,并提议一个以对抗性学习为基础的定位分类器,以自动识别公众对与COVID-19有关的卫生政策的态度。具体地说,我们采用了对抗性学习模式,使该模式能够对大量贴标签的数据进行培训,并从源主题中获取可转移的知识,以便能够以鲜少的标签数据概括到正在形成的卫生政策。与此同时,我们设计了一个地球编码器,鼓励模型学习每个区域指定的未观察到的背景因素,并把它们作为非文字信息来增强模型的更深入理解。我们评估了与COVID-19有关的政策在立体检测任务方面广泛基线的绩效,并实验结果显示,我们拟议的方法在零位目标中都达到了州性业绩目标。